| name | equity-research-update |
| description | Test an existing single-stock pitch against new information (quarterly earnings, investor day, catalyst event, periodic review) and update the original memo in place. Companion to equity-research-customised-process — does NOT build new pitches from scratch. Use when the user says "update [TICKER] after [event]", "post-earnings on [TICKER]", "review [TICKER] after investor day", "is [TICKER] still my view", "refresh [TICKER]", or "periodic review of [TICKER]". For initial coverage of a new name, use equity-research-customised-process instead. |
Equity Research — Update (Companion to Thesis-First)
Test an existing pitch against new information, decide what changed, and update the original memo in place. The original equity-research-customised-process builds pitches from zero; this skill maintains them through the rest of their life.
Output: an updated [TICKER]_pitch.pdf (in-place edit of the original) with a new row in the Changelog appendix. On rare direction reversals, the old PDF is archived and a new primary published.
Core operating principles
Principle 1: Inherit, don't relax, the parent skill's discipline
This skill is downstream of equity-research-customised-process. Everything from that skill's anti-patterns (no fabricated numbers, no fake precision, cited sources only, killing-conditions are sacred) applies here, often more strictly — the post-event moment is exactly when KCs are tested and when fabrication risk peaks.
Principle 2: The three-test result drives the action; no upfront tier
The skill does not ask the user "pick light/medium/heavy effort." The three-test (pillars / killing conditions / load-bearing magnitude) is run first; its result determines the recommended action. User confirms or pushes back. Effort follows from what the new data actually conveys, not from a pre-committed budget.
Principle 3: Update the original PDF in place; archive only on direction reversal
The archive philosophy is one always-current document per pitch. Refinements edit [TICKER]_pitch.pdf directly and append a row to the Changelog appendix. Git history of the markdown source IS the audit trail. Only direction reversals (LONG → PASS/SHORT or pitch closure) trigger archive-and-replace, because at that point it's a fundamentally different view, not an evolution.
Principle 4: Never introduce new pillars via an update
Pillars come from the original Phase 6→8 asymmetry hunt + direction commit. New information can CONFIRM, INVALIDATE, or be SILENT on existing pillars — it cannot spawn a new one through an update. If new info points to a missing pillar, that's a signal to consider a HEAVY rebuild via the original skill, not to back-fill via this one.
Principle 5: Q&A interludes — every step is a checkpoint
Same as the parent skill. After each step, pause with explicit prompt:
"Step N complete. Output: [path]. Questions / edits, or 'continue' to advance."
User-driven cadence; never auto-advance.
When to invoke
Use this skill when there's a specific new information event and you have an existing pitch under coverage:
| Event type | Trigger phrase |
|---|
| Earnings | "update SPOT after Q2 print" / "post-earnings on SPOT" |
| Investor day | "review SPOT after investor day" / "SPOT capital markets day update" |
| Catalyst | "SPOT update on UMG ruling" / "catalyst note on SPOT" |
| Periodic review | "periodic review of SPOT" / "is SPOT still my view" |
Do NOT use this skill for:
- Initial coverage of a name → use
equity-research-customised-process
- Pre-event scenarios (will they beat?) → out of scope; consider Anthropic's
earnings-preview if needed
- Multi-name updates fanned across coverage → not designed for this; run once per ticker
Workflow at a glance
| # | Step | Output | Q&A weight |
|---|
| 1 | Materials gather | updates/[date]_[event]/ raw artefacts | Light |
| 2 | Structured read | updates/[date]_[event]/structured_read.md | Medium |
| 3 | Three-test | updates/[date]_[event]/three_test.md | Heavy |
| 4 | Action follows from three-test | Updated [TICKER]_pitch.pdf in place (+ Changelog row); on reversal, archive + new primary | Heavy |
Each step has a dedicated reference doc. Read them in order:
Files this skill reads from the original pitch
From ~/Claude Projects/Equity Research/[TICKER]/:
| File | Role |
|---|
working/pillars_audited.md | The 2-4 pillars to test against. Pillar IDs (P1/P2/P3) must match across all skill artefacts. |
working/killing_conditions.md | The KCs to test (must be in the structured format introduced by the parent-skill patch). |
working/valuation_outputs.yaml | Current base PT, bull/bear envelope, WACC, FX. Source of truth for any "current PT" reference. |
deliverables/[TICKER]_pitch.md | The memo source — edited in place for refinements. |
deliverables/[TICKER]_pitch.pdf | Rebuilt from the markdown after edits. |
deliverables/[TICKER]_model.xlsx | Updated only if model flex is part of the action. |
The skill does not read or refetch filings/, transcripts/ from prior quarters, industry_brief.md, or the original context.md — those are part of the foundation that the original pitch was built on, and re-reading them is wasted effort for an update.
Files this skill writes
Per update event:
~/Claude Projects/Equity Research/[TICKER]/updates/[YYYY-MM-DD]_[event-type]/
├── materials/ # raw artefacts pulled in step 1
├── structured_read.md # step 2 output
└── three_test.md # step 3 output
And it edits in place:
~/Claude Projects/Equity Research/[TICKER]/deliverables/
├── [TICKER]_pitch.md # edited in place (Changelog row + any thesis/model changes)
├── [TICKER]_pitch.pdf # rebuilt from md
└── [TICKER]_model.xlsx # only if model flexed
And on direction reversal only:
~/Claude Projects/Equity Research/[TICKER]/deliverables/archive/
├── [TICKER]_pitch_v[N]_[date].pdf # old primary moved here
└── [TICKER]_pitch_v[N]_[date].md
Plus an archive-site copy bump:
~/Python/equity-research-archive/tickers/[TICKER]/
├── SPOT_pitch.pdf # replaced with the rebuilt PDF
└── index.html # "Last reviewed: [date]" line bumped; rating + PT updated if changed
Vocabulary discipline (inherited from parent skill)
- Pillar = a testable argument supporting the direction. IDs (P1, P2, P3) are stable across the original memo, the three-test output, and the Changelog appendix.
- Killing condition (KC) = a pre-specified event/number that would prove a pillar wrong. IDs (KC1, KC2, …) stable across
killing_conditions.md and three-test output.
- Load-bearing thesis = the pillar carrying the bulk of the bull/bear edge (in SPOT's case, the GM expansion). Magnitude shift on this pillar matters most.
- Refinement = an in-place edit that keeps the direction unchanged. Updates the original PDF.
- Direction reversal = LONG↔SHORT or any → PASS. Triggers archive-and-replace, not in-place edit.
Never call an update a "new pitch." It's a refinement of an existing one — unless direction reverses, in which case it's a new pitch with the old archived for audit.
What this skill does vs what you do
Same split as the parent skill:
| Skill does | You do |
|---|
| Pulls new artefacts | Decides if more material is needed |
| Reads end-to-end, structures NEW vs RESTATED | Validates the reading |
| Runs the three-test mechanically | Confirms verdicts (CONFIRMED/INVALIDATED/SILENT) |
| Recommends an action based on three-test | Picks the action |
| Drafts the in-place memo edits | Approves / edits the diffs |
| Rebuilds PDF, copies to archive | Confirms the final |
Skill never picks the action; surfaces the recommended one with explicit reasoning. User decides.
