canon-implementation

name	canon-implementation
description	Use when you need a governed implementation run for an existing system with explicit task mapping and mutation bounds.
preflight	{"requires_canon":true,"requires_initialized":true,"canonical_input":"implementation","system_context":"existing","risk_required":true,"zone_required":true,"owner_optional":true}

Canon Implementation

Support State

• available-now
• default visibility: discoverable-standard

Purpose

Expose the delivered Canon implementation workflow as a governed run started from your AI assistant.

When To Trigger

• The user needs a governed implementation packet for an existing system.
• The user already has a bounded execution brief and wants Canon to persist the packet and evidence.

When It Must Not Trigger

• The user still needs to decide the change boundary or preserved invariants; use $canon-change.
• The user is explicitly asking to inspect, approve, resume, or continue an existing run; use the run-scoped skills instead.

Required Inputs

• RISK
• ZONE
• either one implementation brief file, one implementation input folder, or one inline note

Optional:

• OWNER when the user wants to override Git-derived ownership explicitly

Preflight Profile

• Verify canon is on PATH. If missing, point to the install guide.
• Verify .canon/ exists. If missing, point to $canon-init.
• Always bind --system-context existing for this skill.
• Verify risk, zone, and at least one authored input are present before invoking Canon.
• Treat authored inputs under canon-input/ as read-only source material.
• For auto-binding only, treat canon-input/implementation.md or canon-input/implementation/ as the canonical authored-input locations for this mode.
• For a folder-backed packet under canon-input/implementation/, treat brief.md as the authoritative readiness brief and source-map.md as explicit provenance for carried-forward change or architecture context.
• For an explicit inline note, pass it through --input-text instead of materializing a repo file automatically.
• Never infer --input from the active editor file, open tabs, recent .canon/ artifacts, or any other path under .canon/.
• Do not infer readiness by reading prior .canon/ artifacts or published packets directly; the current brief must restate task mapping, mutation bounds, safety-net evidence, and rollback expectations.
• If the user only has a change intent and not a bounded implementation brief, redirect to $canon-change instead of guessing execution bounds.
• If risk is invalid, ask with guided fixed choices: low-impact, bounded-impact, or systemic-impact.
• If zone is invalid, ask with guided fixed choices: green, yellow, or red.

Author Implementation Body Before Invoking Canon

Canon does not invent the implementation body for you. Canon governs, validates, and persists the packet. You (the assistant) MUST author the real implementation body from the bounded source material BEFORE calling canon run --mode implementation.

Do this every time, even when the user starts from a short brief rather than a finished packet:

1. Read the source inputs the user pointed at. Identify the bounded task map, allowed mutation surface, viable implementation options, local ecosystem health, safety-net evidence, adoption implications, and rollback expectations. Do not guess.
2. Compose a single implementation brief at canon-input/implementation.md or a folder-backed packet under canon-input/implementation/. The authored body MUST include all of the following canonical H2 sections with concrete content:
   • ## Task Mapping
   • ## Bounded Changes
   • ## Mutation Bounds
   • ## Allowed Paths
   • ## Executed Changes
   • ## Candidate Frameworks
   • ## Options Matrix
   • ## Decision Evidence
   • ## Recommendation
   • ## Task Linkage
   • ## Completion Evidence
   • ## Adoption Implications
   • ## Remaining Risks
   • ## Ecosystem Health
   • ## Safety-Net Evidence
   • ## Independent Checks
   • ## Rollback Triggers
   • ## Rollback Steps
3. Inline labels or near-miss headings such as Rollback Plan do not satisfy this slice. Use the canonical H2 headings above.
4. Canon preserves those authored sections verbatim in the emitted packet. If a required section is missing or empty, Canon emits ## Missing Authored Body naming the missing canonical heading instead of fabricating filler.
5. If you cannot author a credible implementation body from the available source, stop and redirect to $canon-change or ask targeted clarification questions before invoking Canon rather than submitting an empty brief.

Packet Shape And Persona

Author the packet as an implementation lead comparing bounded execution options for maintainers and approvers inside a task-mapped delivery packet.

• Favor task linkage, options matrix tradeoffs, ecosystem health, adoption implications, decision evidence, and rollback clarity over execution theater.
• Keep recommendations bounded to the declared mutation bounds and allowed paths; do not imply broader implementation authority than the brief supports.
• Persona guidance is presentation only. Missing authored sections remain explicit gaps and must not be backfilled with confident implementation prose.

Canon Command Contract

• Canon command: canon run --mode implementation --system-context existing --risk <RISK> --zone <ZONE> [--owner <OWNER>] (--input <INPUT_PATH> | --input-text <INPUT_TEXT>)
• Return the real Canon run id, state, and any approval target Canon emits.
• Implementation always emits a gate:execution approval target on the first run; the state is AwaitingApproval and the posture stays recommendation-only until that gate is approved.
• After $canon-approve records the gate:execution approval, Canon still remains AwaitingApproval until $canon-resume runs the post-approval continuation.
• After $canon-resume, surface approved-recommendation only when the packet has no executable local patch payload; real workspace mutation currently requires a bounded local payload such as patch.diff inside canon-input/implementation/.

Same-Work Continuity

• Implementation preserves explicit continuation identity continuity, but it does not provide the targeted working-brief lifecycle.
• Candidate detection is advisory. Continuation requires explicit intent.
• Fresh intake stays a new run unless the user explicitly says continue, resume, or same run, or supplies a RUN_ID.
• Use $canon-status for the compact summary and canon inspect refinement --run <RUN_ID> when the user needs the advisory continuation state Canon persisted for the run.
• Do not claim that .canon/runs/<RUN_ID>/artifacts/implementation/working-brief.md exists for this mode unless Canon emits that surface in a future slice.

Expected Output Shape

• concise run-start or gated summary
• Canon-backed run state
• direct statement of the implementation result when a readable packet exists
• primary artifact path and short excerpt when available
• direct statement of the active execution posture (recommendation-only before resume, approved-recommendation only after approved resume without a local patch payload, mutating after approved resume with a valid local patch payload)
• concrete .canon/artifacts/<RUN_ID>/implementation/ paths when Canon emitted them
• one recommended next step that keeps the run context intact
• Action Chips: when the host supports chips, preserve the full objects Canon already returned in mode_result.action_chips; do not collapse them to label-only bullets. In text-only hosts, render each chip's text_fallback instead. Typical gated set: Inspect evidence, Approve generation..., Open primary artifact; after approval but before continuation: Open primary artifact, Inspect evidence, Resume run. Must be the last element of the response; do not place any text after this section.

Failure Handling Guidance

• If canon is missing, show the supported install path from README.
• If .canon/ is missing, point to $canon-init.
• If the user is really still deciding the change boundary, redirect to $canon-change instead of guessing task mapping or mutation bounds.
• If Canon returns AwaitingApproval, surface the exact target Canon produced and do not imply the run is complete.
• If approval is already recorded and Canon still returns AwaitingApproval, say directly that the run now needs $canon-resume, not another approval.
• If Canon returns a blocked packet, point first to the emitted implementation artifacts that name the missing task mapping, mutation bounds, safety-net evidence, or rollback notes.
• If Canon returns recommendation-only or approved-recommendation guidance without a local patch payload, say directly that Canon did not mutate the workspace in this tranche.

Next-Step Guidance

• When Canon emitted a readable packet, recommend $canon-inspect-artifacts first.
• Use $canon-inspect-evidence when the user needs invocation rationale or policy decisions.
• Use $canon-change when the real gap is still bounded planning rather than execution guidance.
• Use $canon-pr-review only when the real target is a diff or WORKTREE.

Related Skills

• $canon-status
• $canon-inspect-evidence
• $canon-inspect-artifacts
• $canon-approve
• $canon-resume
• $canon-change
• $canon-pr-review

Lifecycle Hook Detection

Before starting the governed run, check for before_run hooks:

1. If .canon/hooks.toml exists and is valid TOML with version = 1:

   • Parse hooks.before_run.actions array.

   • Filter by mode_filter: skip hooks whose mode_filter does not include implementation. If mode_filter is null or absent, include the hook.

   • For each matching hook, emit a proposal block:

     ## Lifecycle Hook Detected
     
     **Event**: before_run
     **Hook**: <id>
     **Command**: `<command>`
     **Working Directory**: <repo-root>
     **Description**: <description>
     **Required**: <Yes if optional=false, No if optional=true>
     **Trusted**: <Yes if trusted=true, No if trusted=false>
     
     Proceed with this hook? [Yes / No / Skip all hooks]
     

   • Apply confirmation rules per the optional/trusted matrix in .agents/skills/canon-shared/references/hooks-schema.md.

   • For untrusted hooks (trusted: false), add an extra confirmation step explaining that the command source is not in the trusted allowlist.

2. If .canon/hooks.toml is missing, unreadable, or has invalid TOML, skip hook detection silently. Do not block the run flow.

3. Record hook trace in ai-provenance.md (see Hook Trace Recording below).

Hook Trace Recording

After each hook proposal is resolved, append to the run's ai-provenance.md:

## Hook Traces

| Hook ID | Event | Command | Trusted | Outcome | Exit Code | Timestamp |
|---------|-------|---------|---------|---------|-----------|-----------|
| <id> | before_run | <command> | <yes/no> | <accepted/declined/skipped> | <code or n/a> | <ISO 8601> |