- Home
- Skills
- Prompt & Skill Engineering
- Skill Trigger Doctor
Skill Trigger Doctor
Measure why a skill doesn't fire, rewrite its description with rules proven by before/after runs, and prove the fix.
$7
Works with the AI Agents you already use
Skill Trigger Doctor
Example session with this skill installed
Two of my Claude Code skills never seem to fire. Find out why and fix what you can.
- Read your context and instructions
- Compiled the trigger doctor
Summary
Step 1 (triage) of the Doctor workflow: python3 scripts/skillcheck.py scan, static checks with no model calls. Both skills are switched off in settings, so this is a static problem, not wording. Output below is from a real run, paths shortened.
== claude: 17 skills
! EASY ~/.claude/skills/EASY
[info] Agent Skills spec (agentskills.io) names are lowercase letters, digits and hyphens; other agents may reject this name
! code-ref ~/.claude/skills/code-ref
[info] settings.json skillOverrides is 'off': this skill is turned off
! search ~/.claude/skills/search
[info] settings.json skillOverrides is 'off': this skill is turned off
OK deep-research ~/.claude/skills/deep-research
...
| Skill | Finding | Doctor's next step (references/static-failures.md) |
|---|---|---|
| code-ref | off via skillOverrides | the user turned it off: ask before turning it back on |
| search | off via skillOverrides | same |
| EASY | uppercase name | fine in Claude Code; lowercase-hyphen if shared with other agents |
Next steps
- Once a skill is back on, measure before touching wording:
gena suite with your own phrasings,run --tag before --runs 3. - A miss with "taken by: X" → rule R3 (state the boundary in both skills). A miss with nothing taken → R1 (slash-only wording) then R2 (literal phrasings).
- Re-run on held-out prompts with
--tag afterandcompare; revert anything that made another skill worse.
Connects securely to your tools. The creator never sees your data.
About this skill
For skill authors and heavy users whose skills don't trigger, trigger on the wrong requests, or lose to another installed skill.
What you get
- A guided workflow the agent follows: triage → build a prompt set with your real phrasings → measure the baseline → rewrite → re-run on held-out prompts → before/after report.
- Rewrite rules measured with real
claude -pruns, applied in order. For example: slash-only wording (0/4 → 4/4 on held-out prompts), abstract categories vs literal phrasings (1/4 → 4/4), boundaries stated in both competing skills (2/3 → 3/3). - Description templates for task, gate/review and explain/style skills.
- A guide to reading the report: boundary vs wording vs over-reach problems.
- A preflight step: check a new skill against your installed ones before installing it.
- The measuring tool (skillcheck, MIT, stdlib-only Python) bundled in
scripts/.
Rules the agent follows: change one description at a time, never claim a fix without a re-run, revert any edit that makes another skill worse.
Limits
- Measurement runs on your own Claude Code / Codex subscription, about a minute per prompt run; 5 skills × 6 prompts × 3 runs is roughly 90 minutes in the background.
- The rules come from a limited set of measured cases; they say what to try first, not a guarantee.
- Free version (measurement only): skillcheck on Agensi, or https://github.com/Noem836/skillcheck
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Prompt & Skill Engineering.
Requires Python 3 and your own Claude Code or Codex CLI. Fixes triggering through the description only, not the skill's internal logic. Claude runs load your installed skills and plugins but allow only the Skill tool; Codex runs skip your config, MCP, web and plugins, so Codex-plugin conflicts aren't measured.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 20 days ago
- Accounts
GitHub
- Software
Python 3
Claude Code CLIorCodex CLI
Permissions
Allowed Hosts
Shell: runs python3 scripts/skillcheck.py and your claude/codex CLI. Reads your skill folders; writes suite.json, results and the HTML report in the current folder, and edits a SKILL.md description only after you agree. Network: only if the bundled script is missing (pinned download from GitHub), or a report upload when you run the upload command yourself.