メニュー
How this GTM repository works. Load when orienting in this repo, deciding where a change belongs, or explaining the layer conventions to someone. Covers the layers, the dependency direction, the PR workflow, the confidence ladder, and where each kind of change goes.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Turn something heard in the field into durable context, at the right confidence level. Use when the user relays an objection, a competitor mention, a win or loss reason, a signal, or any claim from a call, and asks to remember it. Enforces the repetition rule and the confidence ladder.
Audit and repair the context layer. Use when asked to clean up context, before a quarterly review, or when the knowledge graph feels stale. Finds missing frontmatter, broken references, orphan files, and stale content, then fixes them file by file.
Write today's entry in the cadence log. Use at the end of a working day, or when the user recaps what happened. Captures what moved, what is stuck, and anything worth remembering, then routes stuck items to carryover and repeated observations toward context.
Scaffold a new GTM motion end to end across all three layers. Use whenever creating a new motion, campaign, play, or automation in this repo, so the narrative (context), the implementation (infra), and the regression suite (evals) ship together and never drift apart. Enforces the outreach hypothesis rule.
Plan the coming week in the cadence layer. Use every Monday, or whenever asked what this week should focus on. Pulls from plan/ and open initiatives/, carries over unresolved items, and writes cadence/weekly/YYYY-Www.md. Escalates anything that has slipped three weeks into a decision instead of a task.
Digest the outputs archive into durable context. Use for the weekly GTM review, or whenever asked what happened recently and what we learned. Reads recent outputs entries, summarizes results, and proposes context updates as a PR. This is the repo's memory flywheel.
| name | gtm-repo |
| description | How this GTM repository works. Load when orienting in this repo, deciding where a change belongs, or explaining the layer conventions to someone. Covers the layers, the dependency direction, the PR workflow, the confidence ladder, and where each kind of change goes. |
This repo follows the Manifest framework (Manifest, by Cargo). Strict dependency direction: infra reads context, skills read everything, nothing depends on outputs.
| You want to... | Touch |
|---|---|
| Set the goal, the forecast, or what we are not doing | plan/ |
| Start a bounded effort with a deadline and success criteria | initiatives/<name>.md |
| Plan the week, log the day, track a dropped ball | cadence/ |
| Update ICP, personas, positioning, objections, proof | context/<domain>/*.md |
| Change what runs in production (models, tools, plays, agents) | infra/**/*.ts |
| Add a repeatable procedure for agents working in this repo | .claude/skills/<name>/SKILL.md |
| Run a one-off operation against the workspace | scripts/*.ts |
| Guard an agent prompt against regressions | evals/<agent>/ |
| Archive the result of a run, campaign, or research task | outputs/YYYY-MM-DD-<slug>/ |
| Keep a half-baked experiment to yourself | scratch/ (gitignored) |
proof/, signal/, and insight/ carry
confidence: hypothesis | validated | proven. Outreach motions may cite only
validated or proven. The linter enforces it. Use the capture-feedback
skill to promote a claim, and only on a second independent occurrence.plan/outcomes.md. Set kind: on every motion.context/motion/. A play is a Cargo resource (definePlay: watches a
data model, runs per row). A motion is implemented as a Cargo play, tool, or
agent. Do not confuse them.plan/outcomes.md names what it becomes in infra/. An
outcome with no Becomes: is a wish.outputs/ entries carry an outcome: field (meetings, replies, pipeline
attributed), so motions are ranked by results rather than opinions.2026-07-14, never "last Tuesday".infra/cargo.state.json is machine-owned. Never edit it or resolve its
conflicts by hand; re-run the deploy instead._template.md. Copy it; do not invent frontmatter
fields.npm run lint (context, outputs, and copy style) if you touched context/,
outputs/, or any prose.npm run typecheck if you touched infra/ or scripts/.npm run eval if you touched a system prompt (needs OPENAI_API_KEY).