agentleFS
EastkapEastkap/autocomp/.claude/skills/initiate-gtm/SKILL.md

initiate-gtm

Skill4 starsChanged 5 months agoMITView on GitHub

Run exactly ONE bounded cycle of the persistent GTM distribution lane — sweep the gtm queue + context stream + scoreboard, atomically claim one gtm-tagged card or run one pre-authorized bot-identity distribution step (directories, SEO/GEO, indexing), verify every submission live, stage gated sends as PENDING, then record and heartbeat. Use when the user says "initiate GTM", "run a gtm cycle", "/initiate-gtm", or the gtm lane runner fires this as its headless cycle prompt.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read roles/gtm.md in full and adopt it.

    line 13
  2. Very specific:

    If private/state/.lane-gtm-pause exists: run tools/context.sh heartbeat gtm paused, report "gtm lane paused (sentinel present)" and STOP — …

    line 19
  3. Distinctive:

    Verify-before-redo, always: check for the live listing/page state FIRST — a crashed predecessor may already have submitted.

    line 29
  4. Distinctive:

    Never self-schedule — the runner owns cadence.

    line 82
  5. Distinctive:

    Draining a whole directory batch in one cycle: instead of decomposing.

    line 92

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

initiate-gtm — one GTM lane cycle

One bounded cycle; the lane runner loops it, you don't. roles/gtm.md IS the contract — this skill adds only the mechanics of a cycle. CLAUDE.md hard rules bind throughout. You sell what already exists — distribution-first is the company's anti-tar-pit rule.

1. Load

Read roles/gtm.md in full and adopt it. Read CLAUDE.md hard rules + private/charter.md (ICP, channels) + the playbook matching the claimed work (tools/directories.md, tools/launchdirectories-list.md, tools/gtm-sources.md, tools/outreach.md, tools/gsc.md; skills: launch-directories, fetch-protected).

2. Gate

If private/state/.lane-gtm-pause exists: run tools/context.sh heartbeat gtm paused, report "gtm lane paused (sentinel present)" and STOP — a cheap no-op cycle, not an error.

3. Sweep (whole queue, never from memory)

  • tools/tasks.sh list-all --tag gtm — ALL statuses.
  • tools/context.sh sweep gtm — unread coordination rows.
  • tools/scoreboard.sh — burn with $0 in = where distribution is owed.

4. Claim & execute SMALL

  • tools/tasks.sh claim gtm — atomic; empty result = empty queue, not an error.
  • Verify-before-redo, always: check for the live listing/page state FIRST — a crashed predecessor may already have submitted; a duplicate submission burns directory reputation.
  • Pre-authorized — do, don't ask: bot-identity GTM under the boseclaw accounts is the standing authorization — directory/launch listings, SEO/GEO/structured data/llms.txt on our own pages, indexing (tools/gsc.py, IndexNow), reading communities, drafting. Re-asking per listing is the timid failure mode.
  • Gated — stage, never execute: any send from the mail domain, mass/cold outbound, spend (paid solves are value-call PENDINGs), anything under the owner's personal identity. Worker-2-held actions (community promo posts, account creation) route to CEO up front.
  • ONE card or ONE playbook step per cycle; a directory batch → decomposed into gtm-tagged cards (plan-big-execute-small). Web tiers: plain HTTP → fetch-protected → Playwright read-only; never classic Chromium for logins.
  • Verify every submission live before recording it: re-fetch the listing URL or read the post-action screenshots; a flow-returned "submitted" is a claim. Unverified stays labeled unverified — no invented listings, ever.
  • Queue empty → tools/registry.sh next, then ONE step of that venture's standing distribution sweep (unsubmitted directories, index coverage, GEO surface, pending-listing checks, outreach drafts staged PENDING). Genuinely nothing viable → tools/context.sh post gtm handoff "<escalation>" --tags ceo — "everything gated" blocks the selling lane, not the company; never falsely idle.
  • 2 stalls on the same step = broken-for-us: route around, kill with the reason recorded, or escalate — never a third identical attempt.

5. Record

  • Finish into review + the qa tag — never straight to done: tools/tasks.sh get <id> → compose the union (the tag verb REPLACES) → tools/tasks.sh tag <id> "<current-tags>,qa" → update <id> review "<what + live evidence>". Listing/ directory cards additionally get the gtm-listing tag in that same union — it is what routes them to the cheap haiku link-check in tools/verifiers.md (without it QA falls through to the generic sonnet-tier check and the cost design silently breaks).
  • Context rows: kind=result per VERIFIED listing with the live URL in refs (--slug <venture> --refs '{"card":"<uuid>","url":"<live>"}'); blocker/handoff as needed.
  • Venture slug worked → private/state/.current-cycle-slug-gtm (one line, overwrite; company for cross-venture work).
  • ONE ledger digest line to private/state/ledger.md via tools/append.sh (pending U5 — until it exists, append with >> only, never editor-style read-modify-write). PENDING gates likewise appended to private/state/approvals.md, never rewritten.
  • Disposition invariant: never exit with your claim in doing — incomplete → todo with progress notes (which directories done, which remain), or decomposed.
  • tools/context.sh heartbeat gtm <ok|idle|blocked> "<one-liner>" — last, always.

6. Report

Final message = the cycle report: per submission — directory, status, live-verification evidence (URL fetched / screenshot read), unverified items labeled as such; drafts staged

  • PENDINGs created; cards moved; what was NOT done this cycle (Principle 11).

Headless deltas (lane runner / claude -p)

  • No AskUserQuestion.
  • Gated classes: PENDING appended to private/state/approvals.md + tools/context.sh post gtm handoff "<gate summary>" --tags ceo (CEO owns the human approval card — you never create human cards) + tools/notify.sh push.
  • Never self-schedule — the runner owns cadence.

Anti-patterns

  • Re-gating pre-authorized bot-identity GTM — a PENDING for a routine listing is the timid failure mode the standing authorization exists to kill.
  • Recording an unverified "submitted" as live — the listing URL fetch IS the record.
  • Re-submitting without checking for the existing listing (reputation burn).
  • Executing a send/spend because it feels small — gates have no de-minimis exception.
  • Marking your own listing done — review + qa tag; QA link-checks it.
  • tag <id> qa without the union — the verb replaces; you just untagged gtm.
  • Draining a whole directory batch in one cycle instead of decomposing.

Cost note

Cheap scoped execution, and the lane Rule 17 exists for — its cadence stays ≥ the CTO's. One claim or one playbook step, live-verify, out. Browser-tier fetches cost seconds, not dollars; paid solves are the only spend and they are PENDINGs, not defaults.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.