alexion/dev
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(skills): simplify craft-skill flows

Browse Source
This commit is contained in:
alexion
2026-09-05 14:05:18 -04:00
parent 4d82c111a2
commit 9b6dabd171
2 changed files with 8 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/skills/craft-skill/GLOSSARY.md
View File

@@ -123,7 +123,7 @@ Hide later steps only when a real context boundary exists.
## Failure modes
Use these findings during audit, import, and verification.
Use these findings during audit and verification.
Quote the offending passage when a finding applies.
### Premature completion

41
src/skills/craft-skill/SKILL.md
View File

@@ -1,12 +1,12 @@
---
name: craft-skill
description: Draft, audit, or import an Agent Skill for structural predictability. Run deliberately as /skill:craft-skill when creating or improving a skill.
description: Draft or audit an Agent Skill for structural predictability. Run deliberately as /skill:craft-skill when creating or improving a skill.
disable-model-invocation: true
---
# craft-skill
Draft, audit, or import an Agent Skill.
Draft or audit an Agent Skill.
Judge the skill against one bar: **predictability**.
Predictability means the agent follows the same process on every run.
@@ -18,7 +18,6 @@ Apply project writing rules when the target project defines them.
Use **Draft** when creating a new skill source.
Use **Audit** when reviewing an existing skill in its authoritative source.
Use **Import** when adapting an external skill into the target project or collection.
Finish when one branch matches the user's requested change.
@@ -47,7 +46,7 @@ Finish when one branch matches the user's requested change.
Give each step a checkable **completion criterion**.
Use relative paths for all skill-local references.
Finish when each branch has a complete path through the skill.
Finish when each branch has ordered steps, resolved references, and checkable completion criteria.
## Audit
@@ -75,39 +74,13 @@ Finish when one branch matches the user's requested change.
Finish when every finding has quoted evidence or no finding remains.
4. Rewrite the authoritative source.
4. Report or fix findings.
Report findings by default.
Fix findings only when the user requests edits.
Preserve behavior that passes the checklist.
Change only structural skill problems unless the user expands the scope.
Keep each retained finding only when a structural reason requires it.
Finish when you have fixed or retained each finding.
## Import
1. Read the source skill.
Read `SKILL.md` and every disclosed file.
Inspect scripts, references, and assets before deciding what to carry forward.
Separate reusable behavior from source-specific packaging.
Finish when the source skill's intent and dependencies are visible.
2. Resolve the target source.
Read the target project's instructions and skill-source conventions.
Choose the authoritative source directory for the imported skill.
Trace enabled-skill links rather than editing them.
Ask the user when the target owner is ambiguous.
Finish when the target path is writable and authoritative.
3. Adapt the skill.
Preserve the source skill's intended behavior.
Replace stale harness assumptions with target-harness mechanisms.
Use valid Agent Skills frontmatter.
Use `disable-model-invocation: true` for Pi user-invoked skills.
Keep useful scripts, references, and assets when they still support the behavior.
Remove source-specific paths, generated-state instructions, and stale packaging rules.
Finish when the target skill expresses one portable behavior.
Finish when you have reported each finding or fixed each requested finding.
## Verify