All skills
simota avatar

/prune

@965f4f9
by shingo imotasimota/agent-skills85 stars
15

Cleaning up the skill ecosystem: auditing the agent roster for overlap and inactivity, proposing merges and sunset plans. Propose-only. Not for ecosystem strategy (Darwin) or code YAGNI (Void).

Use this Skill: https://skilld.dev/gh/simota/agent-skills/prune

This session only. Nothing lands on disk.

referencesunset-protocol.md

≈1.1k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Sunset Proposal and Removal Verification

Prune proposes; the user approves and executes archive/removal. SUNSET and DEPRECATE both require reference/retention-criteria.md's full three-condition gate and protected-skill checks. Low retention does not waive inactivity or alternative coverage. All removal proposals also require reference/pack-impact.md.

Gate and Capability Evidence

gate_evidence:
  condition_1_inactive:
    last_activity: <YYYY-MM-DD or "never">
    days_since: <int>
    pass: <bool>
  condition_2_alternative:
    alternative_skill: <name>
    overlap_with_target: <pct>
    covers_unique_capabilities: <bool>
    pass: <bool>
  condition_3_no_dependency:
    claude_md_mentions: <int with file:line>
    common_md_mentions: <int with file:line>
    pack_memberships: [<list>]
    profile_coverage: [<list>]
    pass: <bool>
gate_verdict: PASS | FAIL (any condition false → FAIL → DEPRECATE-WATCH)

never requires a complete observation window; missing logs cannot prove never-used. Inactivity is ≥6 months, not the 90-day usage-scoring window. Record actual project/host namespaces and all active consumers, not only a hard-coded Claude directory.

capability_migration:
  - capability: <name>
    moves_to_skill: <alternative>
    confidence: HIGH | MEDIUM | LOW
    note: <required prompt phrase shift, if any>

Any capability lacking HIGH-confidence coverage is a partial-coverage gap requiring explicit review; user acceptance is not proof that the alternative condition passes. If a gate fails, record no-action/watch status rather than a ready-to-execute sunset.

Archive and Consumer Manifest

Before proposing the move, search tracked/configured consumers across the full repository and relevant host configuration; preserve command/scope/result. Every discovered consumer joins the manifest below. Archive the skill, its references and associated journal at .archive/<skill>/ with ARCHIVED_<YYYY-MM-DD>.md, reason, alternative and ≥90-day retention/reactivation plan.

Consumer to update Reverse restoration obligation
_common/SKILL_PACKS.md and all actual host profiles Restore recorded Pack/profile membership
Nexus signal-keywords.md and routing-matrix.md Restore keyword targets and default-chain task types
Nexus recipes-index.md and SKILL.md Recipe Registry allowlist Restore both atomically
Nexus agent-disambiguation.md and _common/BOUNDARIES.md Restore disambiguation/boundary entries
Every partner's COLLABORATION_PATTERNS / BIDIRECTIONAL_PARTNERS Restore both directions
Project CLAUDE.md / AGENTS.md, _common and any additional discovered consumer Restore exact references/contract-delivery paths and intentional mirrors/symlinks
Stated skill counts Re-derive from the restored/current roster
.agents/PROJECT.md Append dated archive/reactivation record; don't rewrite history

Reactivation restores the archived directory and this entire recorded consumer set, then notifies Nexus and re-runs repository validation. A directory-only restore is incomplete.

Verification by Absence

After approved execution, repeat the pre-move search. Permitted residuals: .archive/<skill>/, dated .agents/PROJECT.md archive records, CHANGELOG and journal history. A live routing/contract/profile reference anywhere else means not removed. Report actual command and counts; SUNSET_VERIFIED: <N> residual hits, all in archive/records is permitted only after this check, never from a checklist of steps performed.

Proposal Fields and Handoff

Deliver verdict/retention score; all three gate results with dates and file:line evidence; per-capability alternative/confidence; complete downstream consumer/Pack/profile manifest; pre-move residual counts; archive note/path and reverse steps; post-move validation status (pending until execution). Request PRUNE_TO_USER_SUNSET_APPROVAL before any move. After approved execution emit PRUNE_TO_NEXUS_ROUTING_UPDATE with residual-search evidence.

DEPRECATE-WATCH

Persist failed gates in .agents/prune.md; no execution this cycle:

deprecate_watch:
  - skill: <name>
    failed_condition: 1 | 2 | 3
    next_audit_due: <YYYY-MM-DD (90 days out)>
    note: <what would need to change for sunset to be valid>

Source: SKILL.md on GitHub

No alerts13d3 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The prune skill is a maintenance tool for auditing and consolidating an agent's skill roster. It performs read-only analysis of local configuration files and project logs to identify redundant or inactive skills. The skill is designed with a propose-only architecture, ensuring that it never deletes files autonomously and relies on human approval for cleanup actions. No malicious behaviors, exfiltration attempts, or unauthorized command executions were detected.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

Signed by skilld at 965f4f9. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/prune