This is a public repository. Treat every committed file and every Git revision as permanently public, even after a later deletion.
This repository may contain:
- portable
SKILL.mdinstructions and supporting references; - generic Claude Code, Codex, and Hermes adapters;
- redacted examples with fictional identities and placeholder values;
- reusable scripts that read credentials from environment variables or external secret stores.
Do not commit:
- API keys, access tokens, passwords, cookies, private keys, or OAuth credentials;
.env,auth.json,settings.local.json, or secret-store exports;- Hermes or Claude sessions, transcripts, memories, user profiles, runtime databases, snapshots, or backups;
- personal names, chat identifiers, private infrastructure addresses, or machine-specific home paths unless they are clearly fictional examples;
- client data, private project rules, internal URLs, or proprietary project context;
- generated skills copied directly from a local agent without review and redaction.
Private agent state and migration snapshots belong in a separate access-controlled backup. Project-specific rules belong in the relevant private project repository. This repository is only the shared, portable agent layer.
All agent-generated changes should use a branch and pull request. Before pushing:
- Review the complete diff, including generated and untracked files.
- Replace real identities, paths, URLs, and account data with neutral placeholders.
- Keep credentials outside the repository and reference only their environment-variable names.
- Run
./scripts/validate.sh --fulland resolve every validation or secret-scan finding before merge. - Confirm that a new or updated skill remains useful outside the task that produced it.
Do not suppress a scanner finding until the matched value has been inspected and confirmed to be a safe fixture. Prefer changing a realistic credential-like fixture over adding a broad allowlist.
Do not paste a live secret or sensitive data into a public issue, pull request, or discussion.
If a credential was exposed:
- Revoke or rotate it immediately. Deleting the file or commit is not sufficient.
- Check GitHub Actions logs, artifacts, forks, and downstream clones for further exposure.
- Remove the value from the current branch and, when necessary, rewrite affected Git history.
- Re-run secret scanning across the full history.
- Open only a sanitized follow-up issue or pull request that contains no live value.
If no private contact channel is available, rotate the credential first and then open a public issue containing only a redacted description.
Security fixes apply to the current main branch. Historical revisions are retained for audit and should not be treated as supported releases.