agentleFS
mgiovanimgiovani/cc-arsenal/skills/oss-launch/SKILL.md

oss-launch

Skill8 starsChanged 48 days agoMITView on GitHub

Takes a private side project from "done" to a public GitHub launch, pre-flight checks (confirm the repo is still private, scan for leaked secrets, verify a license exists), applies review-code findings, generates brand art (logo/hero via codex-imagegen), rewrites the README and repo description for discoverability, scrubs internal or AI-tooling mentions from code and docs (only after explicit confirmation of what to remove), optionally rewrites git history while the repo is still private, then flips the repo public with topics and description set. Use for "open source this", "get this repo ready to go public", "launch prep", "make this repo public", "prep this for launch", or "clean this up before I open source it". Not for PR/merge mechanics on an already-public repo (use ship) or cutting version releases (use git-release). Not for the actual image-generation call itself (use codex-imagegen directly) or a standalone secrets/license audit with no launch intent (use env-setup).

  • Reads credentials

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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. Distinctive:

    Never silently edit or remove anything here: Ask which matches to remove.

    line 54
  2. Very specific:

    Only reachable if the user explicitly asks for it (a --rewrite-history flag or plain-language request).

    line 58
  3. Very specific:

    "Ever public" is not directly queryable: Treat forkCount > 0 or stargazerCount > 0 from the same gh repo view call as evidence the repo may …

    line 61
  4. Distinctive:

    If any condition fails, report the refusal and move on: don't block the rest of the launch on a history rewrite the user can always do …

    line 65
  5. Distinctive:

    Only run once every earlier stage that applied is reported as passed (or explicitly skipped with a reason).

    line 69
  6. Distinctive:

    Never invent topics or a description that don't match what Stage 4 actually wrote.

    line 76

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

OSS Launch

Take a private side project from "done" to a public GitHub launch in seven staged gates: pre-flight, review, brand, README, mention scrub, history rewrite (optional, hard-gated), public flip. Like ship, every stage reports the real command it ran and stops at the first gate that isn't met: never estimate a result, never skip a gate silently.

Track progress with TodoWrite (one line per stage) so the user sees where the launch train is if a gate stops it.

The stages

#StageGate
1Pre-flightRepo is private; no leaked secrets; a license exists
2ReviewNo unresolved Critical/High findings
3Brand(optional) logo/hero generated or reused
4README + descriptionValue-prop-first README, topics set
5Mention scrubUser confirms exactly what gets removed
6History rewritePrivate only, ever-public check, explicit confirmation this turn
7Public flipEvery earlier stage passed

Stage 1: Pre-flight

  • gh repo view --json isPrivate,licenseInfo,forkCount,stargazerCount -q .: record isPrivate; this is the launch's core assumption and Stage 6 rechecks it independently.
  • Secrets scan: run the same check env-setup's Phase 6 does, grep tracked files and .env/git history for key/secret/password/token patterns (see that skill's scan-patterns.md reference if you need the full pattern set). Any hit stops the whole train: report the finding as file:line only, never the matched value, and go no further.
  • License: check for LICENSE/LICENSE.md/COPYING at the repo root. Missing license is not an auto-fill, ask the user which license they want (MIT, Apache-2.0, etc.) before continuing; don't pick one for them.

Stage 2: Review

Invoke the review-code skill (via Skill, or a Task/Explore→sonnet subagent if that's unavailable) against the default branch. Any Critical/High finding stops the train: report it and let the user decide (fix and re-run, or explicitly accept the risk). Medium/Low findings are reported but don't block.

Stage 3: Brand

Check first whether the repo already has a logo/hero asset (ls common paths: assets/, docs/, .github/, README image references), reuse before generating. If none exists and the user wants one, use the codex-imagegen skill (Codex gpt-6-sol + GPT Image 2.5 Sunburst gpt-image-2.5-sunburst) to generate a logo/hero (via the Skill tool where available, otherwise apply its SKILL.md steps inline), then wire the resulting path into the README. Skip this stage entirely (report SKIPPED: not requested) if the user didn't ask for art and none is needed for the README rewrite.

Stage 4: README + description

Rewrite (don't append to) the README so it leads with the value proposition, not a feature dump. Full checklist (badges, install snippet, demo asset, topics/keyword sourcing) is in references/readme-checklist.md: load it before writing the file. Draw the topics/keywords and repo description from what the codebase actually is (package.json/pyproject.toml/Cargo.toml name+description, detected language/framework), never invent a tagline the project doesn't back up.

Stage 5: Mention scrub

Grep for internal-tooling and AI-assistant mentions across tracked files and commit messages: pattern list in references/mention-scrub-patterns.md, load it before scanning.

  • Present the full match list (file:line or commit ref) to the user, even if it's empty, don't summarize it away.
  • Distinguish a leftover-scaffolding mention (a "Generated with Claude Code" footer, a Co-Authored-By: trailer, a stray internal Slack/Jira reference) from a mention that's part of the product's actual described functionality (e.g. a comment explaining a real Claude API integration the product ships): flag both but don't frame them identically.
  • Never silently edit or remove anything here. Ask which matches to remove; apply only what the user confirms.

Stage 6: History rewrite (optional, hard gate)

Only reachable if the user explicitly asks for it (a --rewrite-history flag or plain-language request). This is the one stage in this skill allowed to force-push, and only under every one of these conditions:

  1. Re-check visibility right now with a fresh gh repo view --json isPrivate (don't reuse Stage 1's result, time has passed). isPrivate: false → refuse outright, explain that a public repo may already have been cloned and rewriting shared history would break those clones. Stop here; no backup, no rewrite command, nothing mutates.
  2. "Ever public" is not directly queryable. Treat forkCount > 0 or stargazerCount > 0 from the same gh repo view call as evidence the repo may have been cloned or seen while public, refuse unless the user explicitly overrides after being shown those numbers.
  3. Explicit confirmation in this turn. A general "yes, launch it" earlier in the conversation does not count: ask specifically, in this turn, whether to proceed with an irreversible history rewrite, and wait for a direct yes.
  4. Run the script, never hand-roll the rewrite. <skill-dir>/scripts/history_rewrite.sh --paths <file> --replace <file> re-checks visibility, backs up, rewrites with git filter-repo, verifies, and prints (never runs) the force-push. Procedure, input file formats and the force-push exception are in references/history-rewrite.md, load it only once conditions 1-3 are satisfied, never before.

If any condition fails, report the refusal and move on: don't block the rest of the launch on a history rewrite the user can always do later while still private.

Stage 7: Public flip

Only run once every earlier stage that applied is reported as passed (or explicitly skipped with a reason). Confirm with the user one last time: this is the point of no return for "nobody's cloned it yet."

gh repo edit --visibility public --accept-visibility-change-consequences
gh repo edit --description "<real value-prop, matches the README>" --add-topic <topic1> --add-topic <topic2>

Never invent topics or a description that don't match what Stage 4 actually wrote.

Reporting

End every run (stopped or completed) with a stage table: # | Stage | Command | Result, one row per stage above, in order. Command is the literal command (or skill invocation) actually run; Result is its actual output or exit status. A stage not reached gets Command: —, Result: SKIPPED: <reason>. Never fill a row from what a stage "should" produce: only from a command that actually ran this session.

Example: stopped at pre-flight (secrets found)

OSS Launch stopped at stage 1/7 — pre-flight.

| # | Stage          | Command                                   | Result                                    |
|---|-----------------|--------------------------------------------|---------------------------------------------|
| 1 | Pre-flight      | grep -iE "(secret|api_key|token)\s*=" .env | .env:14 — possible leaked key (STRIPE_SECRET_KEY) — BLOCKED |
| 2 | Review          | —                                          | SKIPPED: not reached                        |
| 3 | Brand           | —                                          | SKIPPED: not reached                        |
| 4 | README          | —                                          | SKIPPED: not reached                        |
| 5 | Mention scrub   | —                                          | SKIPPED: not reached                        |
| 6 | History rewrite | —                                          | SKIPPED: not reached                        |
| 7 | Public flip     | —                                          | SKIPPED: not reached                        |

Rotate the key and remove it from .env before this can proceed — I won't scan further or touch anything else until the secret is out.

Example: refused history rewrite on an already-public repo

OSS Launch: history-rewrite request refused.

| # | Stage          | Command                                    | Result                                       |
|---|-----------------|----------------------------------------------|------------------------------------------------|
| 6 | History rewrite | gh repo view --json isPrivate,forkCount,stargazerCount | isPrivate: false, forkCount: 2, stargazerCount: 5 — already public, refused |

This repo is already public and has 2 forks — rewriting history now would break those clones. I didn't touch git, and nothing else ran. If there's exposed data, rotate/redact it going forward instead, or say so if you want me to proceed anyway knowing the forks will break.

Example: full launch, completed

OSS Launch complete — repo is now public.

| # | Stage          | Command                                                        | Result                                  |
|---|-----------------|-------------------------------------------------------------------|--------------------------------------------|
| 1 | Pre-flight      | gh repo view --json isPrivate,licenseInfo; secrets grep            | private, MIT license present, 0 secrets    |
| 2 | Review          | Skill: review-code                                                 | 0 Critical/High, 1 Low (non-blocking)      |
| 3 | Brand           | Skill: codex-imagegen (hero.png)                                   | assets/hero.png generated, wired into README |
| 4 | README          | Rewrote README.md                                                   | value-prop lead, install, demo gif, badges |
| 5 | Mention scrub   | grep AI-tooling/internal patterns                                    | 1 match (README credit line) — user confirmed removal |
| 6 | History rewrite | —                                                                    | SKIPPED: not requested                     |
| 7 | Public flip     | gh repo edit --visibility public --accept-visibility-change-consequences | ✓ now public, topics set                |

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.