Build dependency network and critical path from a matter plan. Produce an interactive Gantt, flag near-critical tasks, and run what-if cascade scenarios when delays occur — showing programme impact and drafting communications. Produces filtered workstream or jurisdiction views for local counsel. Trigger on: 'build a timeline', 'Gantt chart', 'critical path', 'what if X is delayed', 'what moves if', 'schedule impact', 'how does this affect the programme', 'when do we finish', 'can we still close on time', 'if we miss this deadline', 'run a what-if', 'visualise the plan', 'Germany timeline', 'what does the Employment workstream look like', 'timeline for local counsel', 'just show me the [workstream] tasks'.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Build dependency network and critical path from a matter plan. Produce an interactive Gantt, flag near-critical tasks, and run what-if cascade scenarios when delays occur — showing programme impact and drafting communications. Produces filtered workstream or jurisdiction views for local counsel. Trigger on: 'build a timeline', 'Gantt chart', 'critical path', 'what if X is delayed', 'what moves if', 'schedule impact', 'how does this affect the programme', 'when do we finish', 'can we still close on time', 'if we miss this deadline', 'run a what-if', 'visualise the plan', 'Germany timeline', 'what does the Employment workstream look like', 'timeline for local counsel', 'just show me the [workstream] tasks'.
Convert a dependency-tagged task list into a visual timeline with critical path analysis. Identify which tasks must complete on time for the matter to close as planned. Model delays before they happen — or calculate their programme-level impact after they do.
The what-if cascade is the central capability. When a jurisdiction delays, a regulatory determination runs long, or a counterparty goes quiet, the question is not just "is this task late" but "which of the twenty-three things downstream of this task now move, by how much, and what does the client need to know?" This skill produces that answer, along with the draft communications.
This skill consumes the structured output of matter-plan-builder and feeds its results back as updated milestone dates to status-report-drafter and scope-change-controller. It sits at the technical centre of the LPM plugin: every other skill produces or consumes timeline data; this skill owns the calculation layer.
Operating Modes
Mode 1 — Baseline build
At matter setup: build the dependency network from the matter plan, calculate the critical path and near-critical paths, produce the Gantt diagram and dependency summary. This is the timing baseline all future reporting measures against.
Input: matter-plan-builder structured export (CSV) or pasted task table with dependency tags. Minimum viable input: Task ID, Duration (working days), Predecessor(s), Dependency type (FS/FF/SS), and at least one fixed date anchor.
Forward-scheduling vs back-scheduling: If a programme start date is known but no fixed end date exists, schedule forward — calculate EF from ES. If a client completion deadline is fixed (the more common scenario in legal work), back-schedule — treat the completion date as the LF of the final task and work backward through the network to derive the required start date for each task. Flag which mode is in use and why. A back-scheduled plan immediately reveals whether the available time is sufficient: if the calculated start date for the first task is in the past, the programme is already in deficit before it begins.
Mode 2 — What-if analysis
At any point during the matter: propose a change and receive a full cascade impact analysis. "German counsel now says 12 weeks, not 10" → this skill calculates which tasks move, which milestones slip, whether the critical path changes, and what the new programme completion date is. Produces a draft client notification and affected local counsel communications.
Input: existing baseline (uploaded or pasted) + proposed change described in plain language or as extracted from email.
Mode 3 — Baseline update
Apply a confirmed change: update the timeline, version it, produce a comparison table (original dates vs revised dates for all affected tasks), and trigger cross-skill handoffs to status-report-drafter and scope-change-controller.
Produce a filtered Gantt and timeline summary for a single workstream or jurisdiction, drawn from the full programme baseline. Use when local counsel needs their timeline without the full programme, when a partner asks for a single workstream update, or when preparing a communication for one team that should not expose programme-wide information.
Input: full baseline (or the relevant subset) + the workstream or jurisdiction to filter on.
Three categories of task appear in a Mode 4 output:
In-scope tasks — tasks belonging to the specified workstream or jurisdiction. Shown as normal coloured bars (critical path red, non-critical green). Float values inherited from the full network — not recalculated on the subset.
Upstream constraints — tasks in other workstreams that are hard predecessors to in-scope tasks. Shown as grey bars labelled "[External — constrains [Task ID]]." The recipient needs to see these to understand why their tasks have the dates they do, but should not be misled into thinking these are their responsibility.
Downstream milestones — programme milestones that in-scope tasks feed into. Shown as milestone diamonds labelled "[Programme milestone — [description]]." These give the recipient visibility of what their work is gating without exposing the full downstream network.
Float must be inherited from the full network calculation, not recalculated on the subset. A Germany registration task with 0 days float in the full programme has 0 days float in the Germany view. If float were recalculated on the Germany subset alone, tasks that are critical in the full network would appear to have float — which would be wrong and potentially dangerous. Always calculate against the full baseline first, then filter.
Before Starting Any Mode
Confirm the following before running any calculation:
Programme start date — the date from which durations are calculated. If not provided, ask.
Fixed external dates — any dates that cannot move: regulatory filing deadlines, contractual completion dates, court dates, end-of-financial-year dates, client board approval windows. These are constraints, not estimates. Flag them explicitly — a fixed date on the critical path changes the calculation entirely.
Duration units — confirm whether durations are working days (Mon-Fri, default) or calendar days. Never infer.
Working calendar exceptions — flag any jurisdictions with non-standard working weeks (e.g. Middle East Friday-Saturday weekends, public holiday-heavy periods). Duration estimates that don't account for local calendars are wrong from day one.
Lag/lead values — where a predecessor has a waiting period before the successor can start (lag: positive days) or where the successor can start before the predecessor finishes (lead: negative days). Common in legal work: "regulatory filing submitted → 6-week determination window before successor can proceed" is a 30-working-day lag on a FS dependency.
Regulatory lags require clarification and three-scenario modelling: when a lag represents an external determination window, confirm whether it is (a) a hard minimum — the authority cannot determine before this period regardless of circumstances, or (b) an expected duration — the typical determination period, which could be shorter or longer.
For expected durations, produce three scenarios as a named output — do not just note the risk in prose:
Best case: minimum credible determination period
Expected: the stated duration
+50%: expected duration × 1.5
Show the programme completion date under each scenario. This is the analysis the LPM needs to have the client conversation about deadline risk before the window opens. If the LPM cannot confirm which type the lag is, model as expected duration, produce all three scenarios, and flag the assumption explicitly.
If critical inputs are missing, stop and ask. Do not calculate against an incomplete network — the output will appear precise while being wrong.
Step-by-Step Process
Step 1: Build the dependency network
List all tasks in topological order — predecessors always appear before successors. Identify:
Network start tasks: tasks with no predecessors (or no in-scope predecessors)
Network end tasks: tasks with no successors — the milestone or task whose completion date is the programme completion date
Dependency edges: for each task, record its predecessor(s) and dependency type (FS/FF/SS) plus any lag/lead value
Lag convention: lag counts forward from the working day after predecessor EF. FS lag 5 WD from EF Wednesday 7 May → successor ES = Wednesday 14 May (8, 9, 12, 13, 14 = 5 WD). Day zero is the day after EF, not EF itself. Apply this consistently — a one-day error in lag convention propagates to every downstream task on the same path.
Flag circular dependencies immediately — they are data errors and prevent calculation. A circular dependency means Task A depends on Task B and Task B depends on Task A. They appear most often when a task's predecessors have been entered as successors by mistake.
If the input contains resource or information dependencies (from matter-plan-builder's dependency register) that do not have their own task row, add them as explicit lag values on the FS relationship — do not silently omit them.
Step 2: Forward pass — calculate Early Start and Early Finish
Work through tasks in topological order. For each task, calculate:
ES (Early Start): the earliest date the task can begin, given its predecessors
EF (Early Finish): ES + Duration (in working days, adjusted for weekends and flagged calendar exceptions)
By dependency type:
FF (Finish-to-Finish): Successor EF ≥ Predecessor EF + lag. The successor cannot finish until the predecessor finishes. Both tasks run in parallel; the constraint controls the finish end, not the start. This is the dominant dependency type in multi-jurisdiction execution work — two parallel workstreams converging at a controlled sequence point. Example: dividend resolution completes Monday; share transfer must complete by Thursday. Both run in parallel throughout; the FF+3d lag controls when the transfer can close. If dividend resolution slips, share transfer slips by the same amount — even though both started simultaneously.
FS (Finish-to-Start): Successor ES = Predecessor EF + lag. The successor cannot start until the predecessor finishes. This is the dependency type for hard legal sequencing requirements — Germany entity registration must complete before Netherlands dissolution can begin; there is no parallel execution possible. Also used with lag for regulatory windows: filing submitted → [30 WD lag] → determination received → successor starts.
SS (Start-to-Start): Successor ES ≥ Predecessor ES + lag. The successor cannot start until the predecessor starts. Used when two workstreams must launch in coordinated sequence but can then proceed independently.
When a task has multiple predecessors, its ES is determined by the latest-finishing predecessor (for FS/SS) or the most constraining relationship across all dependency types. Apply each predecessor constraint independently and take the maximum.
Show the forward pass calculation for the critical path tasks and for any task where the constraint is non-obvious. Transparency builds trust in the output — do not present dates without showing how they were derived.
Step 3: Backward pass — calculate Late Start and Late Finish
Work backwards from the programme end date. The end date is either a fixed constraint (if one was identified in the setup check) or the EF of the last task in the network.
For each task, calculate:
LF (Late Finish): the latest date the task can finish without delaying the programme
LS (Late Start): LF - Duration
By dependency type (backward pass):
FF: Predecessor LF = min(Successor LF - lag) across all successors
FS: Predecessor LF = min(Successor LS - lag) across all successors
SS: Predecessor LS = min(Successor LS - lag) across all successors
When a task feeds multiple successors, its LF is constrained by the earliest-requiring successor. Take the minimum.
Step 4: Calculate float and identify the critical path
For each task:
Total float = LS − ES = LF − EF
Float is the amount of schedule flexibility a task has. A task with 10 days of float can slip by up to 10 days without delaying the programme completion date — assuming none of its successors have already consumed that float.
Zero float = critical path. The task must complete on its earliest finish date to avoid delaying the programme. Zero float does not mean the task is late — it means it has no buffer.
Negative float = already late. The task cannot complete by its latest finish date even if started at its earliest start. This occurs when a fixed deadline constraint is tighter than the network allows. Negative float is a programme-level problem requiring immediate escalation — it cannot be resolved by better scheduling. Flag every task with negative float, name the constraint causing it, and quantify the gap. "T04 has -8 days float: the programme requires completion by 30 June, but the network produces EF of 12 July. An 8-working-day compression is required on the critical path."
All tasks on the critical path is a valid and expected result on small linear matters with no parallel paths. It means there is no float anywhere in the network — every task must complete on time. Note this explicitly: "All tasks sit on the critical path. The network has no parallel paths that could provide float. There is no scheduling flexibility in this plan."
Near-critical path: tasks with float ≤ the near-critical threshold. Default threshold: 5 working days. The threshold is configurable — a fast-moving matter may use 3 days; a long-running programme may use 10. Ask the LPM to confirm or accept the default before producing the near-critical report. Express as: "The following tasks have float within [threshold] working days of the critical path and warrant active monitoring: [list with float values]."
If the critical path passes through a task whose duration is an information dependency (regulatory determination, counterparty response), flag this explicitly: "Programme completion is contingent on [external event]. The LPM has no control over this duration. Build contingency planning around the minimum/expected/maximum range."
Step 5: Produce the baseline output
See Output Format section. Produce in order: Gantt diagram, critical path narrative, near-critical tasks table, dependency summary, open calculation assumptions.
Step 6: What-if cascade (Mode 2 only)
See What-If Cascade Protocol section.
Step 7: Update and version (Mode 3 only)
Apply the confirmed change to the baseline. Increment the version number. Produce the comparison table. Trigger cross-skill handoffs.
Step 8: Produce workstream or jurisdiction view (Mode 4 only)
Run the full network calculation (Steps 1–4) against the complete baseline first. Float values must come from this full calculation.
Identify the filter: which workstream or jurisdiction has been requested?
Classify every task as: in-scope / upstream constraint / downstream milestone (see Mode 4 definitions above).
Produce the filtered Gantt showing all three categories with distinct visual treatment.
Produce a plain-language summary for the recipient: "Your workstream has [X] tasks. The earliest you can start [Task Y] is [Date] — this is gated by [upstream constraint task]. Your work feeds into [programme milestone] on [Date]. Your tasks have [Z] days of float in the full programme."
Note explicitly what is not shown: "This view covers the [Germany] workstream only. [X] other workstreams are running in parallel and are not shown here."
Critical Path Methodology — Key Concepts
Why the critical path matters in legal work:
The critical path is not a scheduling curiosity — it is the answer to "what do I actually have to manage?" Every matter has tasks that can slip without consequence and tasks that, if they slip one day, slip the programme one day. Experienced LPMs develop intuitive critical path knowledge. This skill makes it explicit, which serves two purposes: it survives the LPM's departure, and it produces defensible programme impact assessments when delays need to be communicated to clients.
Effort vs duration:
Duration is the elapsed time to complete a task. Effort is the person-hours applied. They diverge whenever waiting is involved. A two-hour partner review that sits in an inbox for three days before being done has three days duration and two hours effort. In legal work, the waiting periods are frequently longer than the work periods — review cycles, approval queues, regulatory windows. Schedules must be built on duration, not effort. Effort-based scheduling underestimates duration by systematically ignoring the waiting.
Lag and the regulatory window pattern:
The most common critical path driver in cross-border legal work is the regulatory window — a period during which work has been submitted to an external body and the firm is waiting for a determination. This is modelled as a FS dependency with a lag equal to the expected determination period. The difficulty is that this lag is uncertain. Model it as: (a) expected case, (b) +2 weeks, (c) +4 weeks. The what-if cascade handles this automatically — it is the primary use case for Mode 2.
Float is not slack:
Float belongs to the path, not the task. If Task A has 10 days float and feeds Task B which also has 10 days float, and Task A uses all its float, Task B has zero float remaining. Float consumed by an upstream task disappears for all downstream tasks on the same path. This is why a delay "on a non-critical task" can still shift programme completion — the task wasn't critical at planning time, but consumed float that other tasks were relying on.
Near-critical is where monitoring effort goes:
The critical path is managed. Near-critical tasks are where surprises come from — tasks that appeared to have buffer, slipped incrementally, and arrived at zero float without anyone noticing. A near-critical monitoring report, reviewed at every status call, is more valuable than a monthly critical path report that only surfaces problems after the damage is done.
Domain Knowledge — Legal Scheduling Patterns
FF as the primary execution dependency in parallel workstreams:
When multiple workstreams run in parallel and must converge at a controlled sequence point, FF+lag is the correct dependency type — not FS. FS would force one task to wait entirely for the other to complete before it could start, losing the benefit of parallel execution. FF allows both to run simultaneously while enforcing the required sequencing at the finish end.
The canonical example: a dividend resolution must be passed on Monday; the related share transfer must complete by Thursday. Both tasks run in parallel from the start of the week. The dependency is not "transfer waits for resolution to finish before it begins" (FS) — both are underway simultaneously. The dependency is "transfer cannot close until 3 working days after resolution passes" (FF, lag = 3 WD).
This pattern repeats across multi-jurisdiction corporate work wherever parallel workstreams must converge in a defined order: parallel entity preparations converging at a group-level signing; parallel regulatory filings converging at a combined determination; parallel DD workstreams converging at a combined report. In each case, the tasks start together or close to it — the constraint is at the end, not the beginning. Model these as FF+lag, not as FS.
The jurisdiction dependency chain (corporate reorganisations) — FS:
Multi-jurisdiction reorgs have a structural critical path determined by legal sequencing, not resource availability. Germany must complete before Netherlands because the Dutch entity cannot be dissolved until the German parent relationship is registered. Singapore must complete before HK because of holding company sequencing. These are hard FS dependencies — there is no parallel execution possible, because the legal validity of the downstream step depends on the upstream step being complete. The critical path in a reorg is not the longest sequence of tasks — it is the legal completion sequence for the jurisdiction chain.
Signing and execution windows:
Physical signing logistics impose scheduling constraints that do not appear in task lists: wet ink signature requirements, notarisation lead times (3-10 working days per jurisdiction), courier delivery windows, registry queue times. These are not durations — they are fixed calendar events. Model them as milestones with fixed dates, not tasks with estimated durations. A completion that misses the registry queue by one day in a jurisdiction may have a three-week impact if the next registry slot is monthly.
The counterparty silence pattern:
Information dependencies on counterparties behave differently from regulatory windows. Regulatory windows have known minimum and expected durations. Counterparty silence is unbounded and often precedes a substantive problem. Model counterparty response windows as: expected (the agreed review period), +50% (realistic slip), and escalation trigger (point at which the LPM escalates). The what-if cascade should always include the escalation trigger scenario.
Crashing and fast-tracking:
When the baseline shows a programme end date that does not meet the client's deadline, two compression levers are available: (a) crashing — adding resources to shorten critical path task durations; (b) fast-tracking — converting FS dependencies to FF or SS where the work can genuinely overlap without creating rework. Fast-tracking is risky in legal work because many FS relationships are hard legal sequencing requirements, not scheduling preferences. Converting a legal sequencing FS to FF does not make the downstream work legally valid — it produces errors, not savings. Flag clearly which dependencies can be converted (scheduling preference) and which cannot (legal requirement).
Domain Knowledge — Common Scheduling Failures
Building the schedule but not the network. A list of tasks with dates is not a schedule — it is a wish list. The value of this skill is the dependency network. Dates derived from dependency logic are defensible; dates entered as preferences are not.
Missing the regulatory window on the critical path. Regulatory determination periods are frequently omitted from task networks because they are not "firm work" — there is nothing to do while waiting. They must appear in the network as lag values or placeholder tasks because they determine when successor tasks can start. A 6-week regulatory window on the critical path that is not modelled produces a programme plan that is 6 weeks shorter than reality.
Treating all float as available. See the Float is not slack principle above. Reporting "this task has 10 days float" without tracking how much of that float has already been consumed by upstream delays is incorrect and misleading.
Not resetting the baseline after a confirmed change. When a delay is confirmed, the baseline must be updated and versioned. Continuing to report against an outdated baseline produces false variance — tasks appear behind plan because they are behind the old plan, which no longer reflects the agreed programme. Mode 3 exists to force this discipline.
Not knowing the critical path changes. On long matters, the critical path shifts as tasks complete and delays accumulate. A task that was near-critical in Month 1 may be on the critical path in Month 4. The critical path calculation must be rerun after every confirmed change — not just at baseline.
Effort in, duration out. A task assigned to a partner who is also on four other matters, in a week where two days are blocked by client travel, does not have the same effective duration as a task assigned to a dedicated full-time resource. Duration estimates should reflect reality, not theoretical productivity. Flag duration estimates that assume full-time availability when the resources are known to be shared.
What-If Cascade Protocol
This is the highest-value output this skill produces. A well-run what-if cascade takes a single proposed change and converts it into a programme-level impact analysis in plain English — the communication asset the LPM needs to manage the client conversation.
Step 1: Identify the proposed change
What changes? Which task or dependency is affected? By how much? Express as: "[Task ID] duration increases from [X] to [Y] working days" or "[Dependency between Task A and Task B] lag increases from [X] to [Y] working days."
Step 2: Recalculate forward from the affected task
Apply the updated duration or lag to the affected task. Recalculate ES/EF forward through all successor tasks using the dependency types. Do not recalculate backward — that is a separate step.
Step 3: Identify all affected tasks
A task is affected if its ES or EF changes as a result of the proposed change. Record: Task ID, original EF, revised EF, days shifted.
Step 4: Identify affected milestones
Which milestones have predecessor tasks in the affected set? Record: Milestone, original target date, revised target date, days shifted. Milestones are the language of client communication — task-level impacts matter less than milestone-level impacts.
Step 5: Critical path analysis
Does the critical path change? Does the proposed change bring a previously non-critical task onto the critical path? Identify any near-critical tasks whose float drops to zero or below as a result of the change.
Step 6: Programme completion impact
Does the programme end date change? By how much? Express as: "Programme completion moves from [Date A] to [Date B] — a slip of [X] working days / [Y] calendar weeks."
Step 7: Produce the cascade impact table — required output, not optional
Every Mode 2 and Mode 3 output must include this table. Do not substitute prose analysis for the structured table — produce both. The table is the deliverable; the prose is the commentary.
| Task / Milestone | Original date | Revised date | Days shifted | CP status | Workstream |
CP status values: New critical / Remains critical / Now near-critical / Unchanged. Sort by days shifted descending. Flag milestones with a separate row marker. This column is the primary management signal — it tells the LPM where to redirect monitoring attention after the change, not just how much things have moved.
Version history entry (required for Mode 3):
Every baseline update must include a version history entry: "v[X.Y] — [Date]: [Cause of change]. Impact: [programme completion shift]. Source: [email/document reference, sender, date]." Produce as a named section: "Timeline Version History."
Versioning convention: use minor increments (v1.0 → v1.1 → v1.2) for schedule adjustments — duration changes, lag updates, date revisions — where the plan structure is unchanged. Use major increments (v1.0 → v2.0) only for structural changes: new phases added, workstreams removed, dependency network restructured. A Germany registration duration change is v1.1, not v2.0.
Comparison table (Mode 3):
The comparison table lists changed tasks and milestones only. Unchanged tasks are omitted — showing 40 rows of "Unchanged" obscures the actual changes. Sort by days shifted descending. End with a summary line: "[X] tasks and [Y] milestones affected. Programme completion shifts [+/- Z working days]."
Sort by days shifted descending. Flag milestones separately. Highlight tasks that become critical as a result of the change.
Step 8: Draft the communications
Two outputs:
Client notification draft: "We write to update you on a scheduling development affecting the [matter]. [Cause of delay in neutral terms]. The following milestones are now targeted for [revised dates]. We are [action taken to mitigate / confirm no mitigation available]. Please let us know if you would like to discuss."
Affected local counsel notification draft (where applicable): "Please note a change to the programme timeline affecting your workstream. [Specific impact on their tasks]. Please confirm [any required response / adjusted deadline]."
Flag both as DRAFT — the partner reviews before any client communication goes out.
Step 9: Offer recovery options
After presenting the cascade impact, offer to model recovery scenarios. Do not model them silently — the LPM and partner decide whether to pursue compression. Ask: "Do you want me to model recovery options? Available approaches for this matter: (a) [Crash specific tasks — name them and the resource implication]; (b) [Fast-track — convert [specific FS dependency] to SS or FF if the work can genuinely overlap]; (c) [Defer [specific downstream task] to a later phase — flag as scope signal for scope-change-controller]. For each option, I can calculate the revised programme completion date and what it would require."
Flag whether each option is genuinely available: not all FS dependencies can be converted, and not all tasks can be resourced up without creating rework risk. Do not offer crashing or fast-tracking as options where the dependency is a legal sequencing requirement rather than a scheduling preference.
Output Format
All outputs produced as .docx unless the user explicitly requests otherwise. These are matter records — they belong in the matter folder.
Every output includes the identifier block. If client name, matter name, or matter number have not been provided, stop and ask before producing output — do not proceed with placeholders silently:
REQUIRED: Produce an interactive HTML canvas Gantt for every Mode 1, 2, and 3 output. This is the primary visual output — not optional, not on request. Do not produce a text table as the visual. Do not produce Mermaid as the default. Render the Gantt inline using HTML canvas. If the rendering fails, note the failure explicitly and produce the text table as a fallback — do not silently substitute.
The HTML Gantt must include:
Colour-coded task bars: critical path in red (#E24B4A), non-critical in green (#1D9E75), regulatory/external wait periods as dashed grey bars
Float shown as a faded bar extension beyond the task bar, with the float value in days labelled
FF dependencies shown as a bracket connecting the finish points of the two parallel tasks, with the lag value in a pill label — not as an arrow
FS dependencies shown as directional arrows between bar ends and starts
Milestone diamonds at programme completion and phase gates
Horizontal axis with working week numbers and calendar dates
What-if slider where a regulatory window or counterparty delay sits on the critical path
For export to a status report or client deck: screenshot the rendered Gantt. A screenshot of a properly rendered HTML Gantt is better than anything built natively in PowerPoint or Excel.
Fallback — Mermaid gantt:
Use only when the user explicitly requests a lightweight output for pasting into a document, or when the matter has fewer than 8 tasks. Note FF/SS relationships in the dependency summary table — Mermaid only renders FS natively.
Near-critical tasks table:
| Task ID | Task Summary | Workstream | Float (working days) | Predecessor at risk |
Cascade impact table (Mode 2/3):
| Task / Milestone | Original date | Revised date | Days shifted | CP status | Workstream |
CP status values: New critical (task moves onto critical path), Remains critical, Now near-critical (task moves off critical path into near-critical band), Unchanged. The status change column is the management signal — it tells the LPM where to redirect monitoring attention, not just how much things have moved.
Workstream view output (Mode 4):
The filtered Gantt uses three distinct visual treatments:
In-scope tasks: normal coloured bars (red = critical path, green = non-critical), float extension shown
Header on the filtered output: "[Workstream/Jurisdiction] view — extracted from full programme baseline v[X.Y]. Float values reflect position in the full programme network, not this view alone. [X] other workstreams not shown."
Do not produce a critical path narrative for a workstream view — the critical path is a property of the full network, not the filtered subset. Instead produce: "Your workstream summary: [X] tasks, earliest start [Date], feeds programme milestone [description] on [Date], [Y] days float in the full programme."
Structured data export:
Every timeline output is accompanied by a CSV export containing all tasks with their calculated ES, EF, LS, LF, float, and critical path flag. This is the input format for scheduling software or SharePoint tracking. Produce inline as a labelled section if a file cannot be attached.
Calculation assumptions (required):
Every output must close with a named section: "Calculation Assumptions." List: working calendar applied, any duration estimates flagged as uncertain, any information dependencies with assumed lag values, any constraints (fixed dates) that override the calculation. Assumptions that are not documented cannot be challenged and cannot be updated when circumstances change.
Cross-Skill Handoffs
From matter-plan-builder: Dependency-tagged task and milestone list (Mode 2/3 structured export) is the primary input. Consume the 13-field task schema directly — do not reinterpret as prose. The matter-plan-builder handoff prompt should read: "Build a dependency network and critical path from this plan."
To risk-and-issues-manager: When a Mode 2 or 3 change confirms a delay, log it as an Issue in the RAID log (confirmed, not a risk). Separately, if the delay is driven by an assumption that may be compressible (e.g. a process-assumption lag rather than a regulatory requirement), log this as a Risk with the compression question as the review trigger. Pass with: "Germany registration delay confirmed — log as Issue. FF lag assumption may be compressible — log as Risk pending counsel confirmation."
To status-report-drafter: Revised milestone dates and critical path status are the reporting baseline. When a Mode 2 or 3 output changes milestone dates, pass the updated milestone register to status-report-drafter with: "Update the milestone reporting baseline — revised dates below."
To scope-change-controller: If a Mode 2 what-if analysis reveals that the scope of work must change (phases added, jurisdictions dropped, work deferred) to recover the programme, flag this as a scope signal. Pass with: "Proposed programme recovery involves scope changes — scope-change-controller should assess OOS implications."
To matter-plan-builder: If Mode 2 analysis confirms that the plan structure must change (new tasks added, workstreams restructured), trigger a Mode 5 plan update. Pass with: "Cascade analysis confirms these plan changes — matter-plan-builder Mode 5 to apply."
To stakeholder-comms-planner: Revised milestone dates and draft communications from Mode 2 cascade. The comms planner handles routing and timing; this skill drafts the content.
From risk-and-issues-manager: When an assumption in the RAID log is breached — particularly an information dependency on a regulatory body or counterparty — that is a Mode 2 trigger. Pass the RAID entry with the assumption breach flagged.
Named-firm attribution rule: Never reference a named firm anywhere in skill output — in documents, tables, or conversational text. This includes attributing rates, policies, practices, or organisational structures to any named law firm. The skill does not know any firm's actual structure, rates, or policies. Use "confirm with Pricing", "confirm with Finance", or "firm policy — confirm before applying." The rule applies to everything this skill produces, not just formal documents.
M365 Connected Mode (Optional)
Connected mode invocation rule: Search connected systems (Outlook, SharePoint, Teams) when doing so adds value — not as a default first step when sufficient input is already in the prompt.
Sufficient input already provided: User has pasted emails, documents, or data with full context. Engage with what is there. Do not search first — it adds friction without adding information.
Input is incomplete or proactive surfacing is warranted: User references something that should be retrieved ("there's an invoice in Outlook", "it's end of month"), or connected mode is running in background/scheduled mode. Search proactively — this is the inverted invocation model and is the highest-value connected mode behaviour.
The distinction is whether the user has already provided what is needed. If yes, work with it. If no, or if proactive surfacing serves the LPM, search.
When the M365 MCP connector is enabled (Claude Team/Enterprise), this skill can:
Pull the approved matter plan from SharePoint (the structured export produced by matter-plan-builder) directly, without the LPM having to upload it
Search Outlook for jurisdiction updates that signal timeline changes — flagging emails where a counterpart mentions delays, revised timelines, or rescheduled actions as Mode 2 triggers
Update milestone dates in a SharePoint List or Planner when a Mode 3 baseline update is confirmed
Create or update calendar reminders for milestone due dates and phase gate reviews
Draft the client notification and local counsel communications in Outlook draft mode for partner review
Without the connector: upload the matter-plan-builder CSV export, paste the relevant email text, and confirm changes directly in the chat.
