agentleFS
alfonsusenricoalfonsusenrico/pakem/skills/sensitive-operations/SKILL.md

sensitive-operations

Skill0 starsChanged 9 days agoMITView on GitHub

Use before touching secrets, .env or credential files, deploy keys, permissions, production systems, deployments, database migrations, or deleting or overwriting data or files. Defines authority, keys-only secret inspection, migration safety, and destructive-target checks.

  • Reads credentials

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 11 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use least privilege and the project's existing approved access paths.

    line 14
  2. Distinctive:

    Never bypass sandboxes, permissions, hooks, or security checks to make a failing operation pass.

    line 15
  3. Distinctive:

    Never print, log, commit, or paste credentials, tokens, API keys, passwords, private keys, or sensitive personal data.

    line 20
  4. Very specific:

    Prefer configuration schemas and .env.example when checking required names.

    line 27
  5. Distinctive:

    Never source or execute an environment file to inspect it.

    line 28
  6. Distinctive:

    If a secret is exposed (printed, logged, committed, or sent to a model), stop, tell the owner which secret and where, and recommend …

    line 29
  7. Distinctive:

    Treat user data, uploads, embeddings, traces, prompts, and model output as disclosure channels.

    line 33
  8. Very specific:

    Never use a personal machine identity (for example ~/.ssh/id_ed25519) for CI, pipelines, or automated deployment.

    line 38
  9. Very specific:

    Create a dedicated key pair per project and purpose (for example ~/.ssh/<project>-deploy), authorize only its public key on the target with …

    line 40
  10. Distinctive:

    Never connect to or act on production without the owner's explicit instruction for that action.

    line 48
  11. Distinctive:

    Follow migration safety: versioned migrations, compatibility windows, ordering, recovery and rollback, mixed-version operation, and …

    line 57
  12. Distinctive:

    Never target a home directory, workspace root, filesystem root, or other broad collection recursively.

    line 66
  13. Distinctive:

    Prefer recoverable operations (move to a backup, soft delete) and keep a recovery path.

    line 68
  14. Distinctive:

    Ask before deleting data or files the current task did not create, and whenever scope or recoverability is uncertain.

    line 70

≈ “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

Sensitive and Production Operations

1. Authority

  • Inspecting, modifying, deploying, migrating, publishing, and deleting are separate authorities. Permission for one does not grant the others.
  • A configured connection, available credential, skill, or API key is not itself authorization to use it.
  • Use least privilege and the project's existing approved access paths.
  • Never bypass sandboxes, permissions, hooks, or security checks to make a failing operation pass.

2. Secrets and sensitive data

  • Never print, log, commit, or paste credentials, tokens, API keys, passwords, private keys, or sensitive personal data.
  • To learn which variables a secret-bearing file defines, run the keys-only reader: python3 scripts/env_keys.py <file> or python3 scripts/env_keys.py .env --compare .env.example. It prints validated names only and never values, comments, unparsable lines, or multiline continuations. Do not improvise text tools for this.
  • Prefer configuration schemas and .env.example when checking required names.
  • Never source or execute an environment file to inspect it.
  • If a secret is exposed (printed, logged, committed, or sent to a model), stop, tell the owner which secret and where, and recommend rotation. Do not repeat it.
  • Applications consume secrets through their normal runtime mechanism; agent investigation must not reveal them.
  • Treat user data, uploads, embeddings, traces, prompts, and model output as disclosure channels. Follow logging and privacy.

3. Deploy and automation identity

  • Never use a personal machine identity (for example ~/.ssh/id_ed25519) for CI, pipelines, or automated deployment.
  • Create a dedicated key pair per project and purpose (for example ~/.ssh/<project>-deploy), authorize only its public key on the target with the minimum access, and store the private key only in the CI secret store.
  • Validate keys only through safe metadata: path, ownership, permissions, fingerprint, public key. Never print private-key contents.

4. Production and release

  • Never connect to or act on production without the owner's explicit instruction for that action.
  • Reproduce and fix locally when practical, then release through the approved pipeline. No manual edits to production code, environment files, or schema.
  • Tie release and runtime evidence to the exact deployed revision and artifact.
  • Report unresolved risks before publishing, deploying, or migrating.

5. Migrations

Follow migration safety: versioned migrations, compatibility windows, ordering, recovery and rollback, mixed-version operation, and data-preservation evidence before calling a migration complete.

6. Destructive operations

  • Confirm deletion or overwrite is within the requested scope.
  • Resolve and inspect exact targets first. Never rely on unresolved variables, broad globs, or command substitution to identify them.
  • Never target a home directory, workspace root, filesystem root, or other broad collection recursively.
  • Prefer recoverable operations (move to a backup, soft delete) and keep a recovery path.
  • Ask before deleting data or files the current task did not create, and whenever scope or recoverability is uncertain.
  • Afterwards, report exactly what was removed and whether it can be recovered.

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.