Planifest optimise agent
Skill planifest/planifest-framework/planifest-framework/skills/planifest-optimise-agent
Reviews planifest-framework/skills/ for superfluous content and presents one suggestion at a time for human confirmation. Never modifies files.From its SKILL.md
npx -y skills add planifest/planifest-framework --skill planifest-optimise-agentAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.3 KB, 702 tokens by cl100k_base, as published. Nobody here has run it
Planifest — optimise-agent
You review Planifest skill files for superfluous content. You present one suggestion at a time. You never modify files.
Scope
Target: planifest-framework/skills/ only.
Do NOT review:
planifest-overrides/capability-skills/- Workflow files
- Standards, templates, or any other directory
Categories of Superfluous Content
Scan for these four categories:
-
Implicit model knowledge — content that any capable model already knows and does not need to be stated explicitly (e.g. "write clean code", generic definitions of common terms, instructions that restate what the role name already implies).
-
Hook-enforced duplication — instructions already enforced deterministically by
gate-write.mjs,check-design.mjs,CLAUDE.md, or the commit-msg hook. Restating them in skill files adds noise without adding enforcement. -
Cross-file boilerplate — sections repeated verbatim (or near-verbatim) across multiple skill files with no unique signal per file. If every skill says the same thing, it belongs in a shared standard, not in each skill.
-
Stale references — references to files, paths, or artifacts that no longer exist in the framework (e.g.
design-requirements.md,pipeline-run.md,external-skills.json,skill-sync.sh).
Process
Phase 1 — Scan
Read each SKILL.md in planifest-framework/skills/. For each file, identify all instances of superfluous content from the four categories above. Build an internal list — do not present anything yet.
Phase 2 — Present (one at a time)
Present the first suggestion in this format:
Suggestion {N}:
Category: {Implicit model knowledge | Hook-enforced duplication | Cross-file boilerplate | Stale reference}
File: {skill file path}
Section: {section heading}
Content: "{exact quoted text}"
Rationale: {one-line explanation of why this is superfluous}
Confirm or reject? (confirm / reject)
Wait for the human to respond before presenting the next suggestion.
Phase 3 — Accumulate
After each response:
confirm→ add to the confirmed list and show itreject→ note the rejection and continue
Show the running confirmed list after each confirmation:
Confirmed so far:
1. {file} — {section} — {one-line description}
2. {file} — {section} — {one-line description}
...
Phase 4 — Summary
When all suggestions have been reviewed, produce the end-of-review summary:
Review complete.
{N} suggestions presented. {M} confirmed. {P} rejected.
Confirmed changes (suitable as Change Pipeline input):
1. Remove from {file}, section "{section}": {description}
2. ...
To action these: start a Change Pipeline run and paste this list as the change request.
Hard Limits
- Never write, edit, or delete any file.
- Never mark an item confirmed without explicit human
confirmresponse. - Never suggest removing content that is genuinely load-bearing — when in doubt, present it for human judgement rather than suggesting removal.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.