agentleFS
ImYourBoyRoyImYourBoyRoy/agent-continuity-stack/skills/apply-project-updates/SKILL.md

apply-project-updates

Skill1 starsChanged 4 months agoMITView on GitHub

Apply an authorized dependency, runtime, toolchain, CI, container, SDK, or tracked-tool update in reversible batches. Use after targets are selected; do not audit broadly or upgrade everything automatically.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    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:

    If the request says only “update everything,” run or request a bounded audit before mutation.

    line 20
  2. Very specific:

    Use references/batching-and-rollback.md to design the batch.

    line 32
  3. Distinctive:

    Use the repository's active package manager and supported update mechanism.

    line 36
  4. Distinctive:

    Prefer exact component updates over broad resolver churn.

    line 38
  5. Distinctive:

    Never pipe an unaudited remote script into a shell.

    line 40
  6. Distinctive:

    Do not install global tools or modify the host toolchain unless separately authorized.

    line 41
  7. Distinctive:

    Do not commit, push, deploy, publish, release, or open a pull request unless that external action is separately authorized.

    line 43
  8. Distinctive:

    Keep security response, compatibility work, routine maintenance, and major migrations in separate batches.

    line 47
  9. Very specific:

    Read references/ecosystem-application.md for the detected ecosystem.

    line 50
  10. Distinctive:

    After each mutation, inspect:

    line 54
  11. Distinctive:

    Stop and explain unexplained source changes, mutable references, registry changes, or excessive resolver churn.

    line 65
  12. Distinctive:

    Run the smallest project-approved checks that exercise the batch.

    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

Apply Project Updates

Implement only the approved update set and preserve a defensible rollback path.

Confirm the selected batch

Require:

  • exact components or an explicitly selected audit batch
  • target versions, tags, or immutable commits when known
  • compatibility and migration evidence
  • repository-approved validation commands
  • authorization for manifest, lockfile, toolchain, or infrastructure changes

If the request says only “update everything,” run or request a bounded audit before mutation. Do not silently choose every available major version.

Capture the baseline

  1. Resolve the repository and applicable instructions.
  2. Inspect Git status and preserve unrelated work.
  3. Record current manifests, lockfiles, toolchains, generated files, and relevant validation results.
  4. Identify the files and artifacts each approved updater may rewrite.
  5. Establish the rollback mechanism before running a mutating command.

Use references/batching-and-rollback.md to design the batch.

Apply through project authority

  • Use the repository's active package manager and supported update mechanism.
  • Preserve one lockfile and do not switch package managers.
  • Prefer exact component updates over broad resolver churn.
  • Inspect installer or migration scripts before execution.
  • Never pipe an unaudited remote script into a shell.
  • Do not install global tools or modify the host toolchain unless separately authorized.
  • Do not commit, push, deploy, publish, release, or open a pull request unless that external action is separately authorized.
  • Respect remote-only builds, hardware requirements, staging gates, and CI ownership.
  • Keep security response, compatibility work, routine maintenance, and major migrations in separate batches.

Read references/ecosystem-application.md for the detected ecosystem.

Inspect the delta

After each mutation, inspect:

  • direct and transitive version changes
  • added, removed, yanked, or source-switched packages
  • changed checksums or integrity records
  • runtime, platform, or compiler requirements
  • generated configuration or bindings
  • feature and default changes
  • migration edits
  • unexpected unrelated manifest or lockfile churn

Stop and explain unexplained source changes, mutable references, registry changes, or excessive resolver churn.

Verify before continuing

Run the smallest project-approved checks that exercise the batch. Broaden based on risk. Do not begin the next batch while the current one has unexplained failures.

Classify:

  • passed and ready
  • passed with existing warnings
  • failed because of the update
  • blocked by environment or policy
  • compatibility uncertain

On failure, repair only when the migration remains within the approved scope. Otherwise restore the batch's own changes without discarding unrelated work.

Report

Provide:

  • approved batch and resulting versions
  • commands and files changed
  • dependency-graph and lockfile delta
  • migrations applied
  • validation and environments
  • warnings, failures, and deferred updates
  • rollback performed or retained
  • next batch requiring approval

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.