agentleFS
fabioc-alohafabioc-aloha/Alex_ACT_ONE/skills/bootstrap-core/SKILL.md

bootstrap-core

Skill0 starsChanged 9 days agoMITView on GitHub

Activates, verifies, repairs, or removes this plugin's user-scope runtime instructions from canonical installed sources. Use after installing or updating Alex ACT ONE, when ACT behavior or Alex Finch identity is inactive, or when bootstrap receipt hashes drift.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 8 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Verify its inclusion in the selected installed package and activation preview.

    line 17
  2. Very specific:

    Run the bundled script without --apply:

    line 24
  3. Very specific:

    When COPILOT_HOME is an absolute path, the default target is $COPILOT_HOME/instructions.

    line 37
  4. Distinctive:

    Start with a host-specific readiness report when the activation steps are unclear:

    line 55
  5. Very specific:

    After an explicit yes, rerun the same command with --apply.

    line 84
  6. Very specific:

    Do not silently resolve a conflict or add an overwrite flag.

    line 97
  7. Distinctive:

    Do not install, update, enable, disable, or uninstall plugins here.

    line 150
  8. Distinctive:

    Do not copy plugin directories, caches, settings, or unowned instructions.

    line 151
  9. Distinctive:

    Do not write unowned instructions without the separately previewed and approved conflict override.

    line 152
  10. Distinctive:

    Do not fetch instruction bodies from the network.

    line 154
  11. Distinctive:

    Do not silently apply during install or session start.

    line 155
  12. Distinctive:

    Do not delete by filename glob.

    line 156

≈ “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

Bootstrap Core

Activate this package's legacy instruction delivery with nothing else installed. Current Copilot clients document native plugin rules, but that is not proof this legacy package's instructions are active. Until the native-delivery compatibility gate passes on supported hosts, this skill copies only manifest-declared canonical instructions and their receipt after explicit consent.

The session-health instruction carries the shared artifact storage contract. Verify its inclusion in the selected installed package and activation preview. A canonical-source edit does not update an installed cache or user profile; installation and activation remain separately approved steps. Instruction parity is not proof of host-enforced delivery.

Preview First

Run the bundled script without --apply:

node <this-skill>/scripts/bootstrap-core.cjs

The preview resolves the canonical sources from the installed plugin root, verifies exact parity with the plugin manifest's instruction inventory, calculates source and destination hashes, reports the resolved distribution target and its source, reports create, replace, or preserve actions, includes the installed plugin version, inspects legacy mixed-receipt evidence, and writes nothing.

When COPILOT_HOME is an absolute path, the default target is $COPILOT_HOME/instructions; otherwise it is ~/.copilot/instructions. Use --target-instructions <path> only for an explicit alternate target or a disposable test. Use --workspace-instructions <path> to recursively report possible workspace overlap before applying user-scope instructions.

The preview also exposes routing.canonicalRoot and routing.skillRoutes. Resolve instruction-to-skill links against the canonical installed source, not the activated profile directory. Missing, undeclared, or externally redirected skill owners block activation before any writes. Native skill discovery remains the preferred loading mechanism; the routing map is a read-only fallback, not evidence that a skill was invoked.

Instructions activate per app, not per machine. Each app that sets its own COPILOT_HOME gets its own target, so activation runs once per app.

Host Readiness

Start with a host-specific readiness report when the activation steps are unclear:

node <this-skill>/scripts/host-readiness.cjs --host scout
node <this-skill>/scripts/host-readiness.cjs --host scout --json

Supported hosts are copilot-cli, vscode, github-copilot-app, and scout. The report runs the bootstrap preview for the selected target and names what still belongs to that host. It writes no instructions or MCP registry entries, does not enable UI settings, and cannot prove a host has restarted or observed the activation.

For Scout, the report also previews its separate MCP registration and names the required Load Copilot CLI skills setting and full restart. Review the existing bootstrap and registrar previews, then run each --apply command separately after consent.

Apply After Consent

Show the resolved target, its source, exact file actions, user scope, receipt action, overlap report, and installed plugin version. Ask:

Activate these instructions for every workspace on this machine?

Quote the actual count from the preview's expectedFiles rather than a number remembered from last time.

After an explicit yes, rerun the same command with --apply. The script writes only changed files, writes .alex-act-one-bootstrap.json atomically when its content needs creation or refresh, and verifies every destination and the receipt against canonical sources. A no-op apply preserves receipt bytes.

Instruction conflicts

The default preview reports blocked: true and names conflicts when a matching destination is unowned or differs from both its receipt and the current source. --apply then fails before writing any instruction or receipt. Host readiness also reports blocked, not ready-to-apply. Unchanged receipt-owned older instructions can still upgrade normally.

Do not silently resolve a conflict or add an overwrite flag. Back up the affected files and receipt, explain which custom bytes would be lost and which unowned files would become receipt-owned, then preview the same target with --overwrite-conflicts but without --apply. Only after explicit approval of those exact replacements may the same command use both flags. This override cannot be combined with removal and does not bypass receipt validation or stale-plan checks.

The receipt owns only installed or explicitly adopted instruction files; ordinary activation never claims a user-authored file merely because its name matches. A legacy mixed receipt is evidence, not authority to rewrite another plugin's state.

Receipts written through version 0.3.0 record alex-act-core as bootstrappedBy and as each entry's owner, with a former source path. A subsequent --apply validates their hashes and migration shape, then refreshes them to alex-act-one ownership and the canonical instructions/ source path without rewriting unchanged instruction files. Preview reports the legacy receipt as a refresh; --remove continues to recognize it as owned until the migration runs.

Repair And Idempotency

A current bootstrap requires:

  1. Canonical source instructions matching the manifest inventory exactly.
  2. A schema-v2 receipt carrying the installed plugin version.
  3. One disjoint receipt entry per current instruction, retaining valid owned retired entries so a later consented removal can clean them.
  4. Source, destination, and receipt SHA-256 parity.

Equal versions do not hide byte drift. A second preview after apply must report only preserve actions and a preserved receipt. An I/O failure may leave partial canonical writes, but must not publish a success receipt or leave temporary write files. Stop, inspect a fresh preview, and use the separately consented conflict-recovery flow when ownership is incomplete; do not infer ownership from a partially written file.

Remove Instructions

Preview removal with --remove. Apply only after explicit removal consent by adding --apply. The script removes only receipt-owned destinations whose current hashes still match the receipt. Modified or unowned files are preserved and reported. Current entries must match manifest-backed ownership; retained retired entries must satisfy the same owner, name, source-path, and hash rules. Unsafe, duplicate, foreign, or malformed entries fail closed before path resolution. Valid retired entries remain owned across upgrades. Clean removal verifies every deletion and removes the receipt. Modified owned bytes and their receipt remain as recovery evidence. Files owned by another plugin's receipt are never removed.

Boundaries

  • Self-activation is not general plugin lifecycle management.
  • Do not install, update, enable, disable, or uninstall plugins here.
  • Do not copy plugin directories, caches, settings, or unowned instructions.
  • Do not write unowned instructions without the separately previewed and approved conflict override. Never write external continuity state here.
  • Do not fetch instruction bodies from the network.
  • Do not silently apply during install or session start.
  • Do not delete by filename glob; receipt and hash ownership are required.

Anti-Patterns

Anti-patternCorrection
Require another plugin before this one can activateRun this plugin's own bootstrap command
Copy instructions without previewShow exact actions and machine-wide scope first
Treat a skill as equivalent to always-applied policyActivate the canonical instruction files
Delete every alex-act-* fileRemove only valid receipt entries with matching hashes
Claim a greeting instruction this plugin did not installKeep separate receipts per owning plugin
Assume one activation covers every appRun activation in each app that sets its own COPILOT_HOME

Would Revise If

Revise if this plugin alone cannot activate every declared source, a preview mutates state, a receipt claims a file it does not own, source resolution fails in a delivered plugin, or removal deletes modified bytes.

Compatibility command constraints: bootstrap-core

Preview exact profile actions first. Apply only after explicit activation/repair consent; removal needs its own explicit consent. Block foreign or modified owned files by default; conflict replacement requires a separate backup, preview, and explicit 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.