agentleFS
Sign inStart free
theopenappstheopenapps/plugin-stack/skills/audit-discoverability/SKILL.md

audit-discoverability

Skill4 starsChanged 10 days agoMITView on GitHub

Audit whether agents would find and pick an AI plugin or MCP server when a user asks: how its listing and MCP tool descriptions answer target prompts, with a labelled simulated pick against rivals. Use when asked to audit a plugin or MCP server, when agents ignore a plugin, to check prompts people should use it for against rivals, for AEO for plugins, to grade a pasted tool list or tools against a rival's, to check whether an app can be reached in a country, or to fix MCP tool descriptions and annotations such as readOnlyHint. Not for writing store copy.

The file

Create a free account to mount
Files
SKILL.md

Audit discoverability

A question such as "how do I figure out what's wrong?" or "what do I need?" is a request to run these steps on this repo, so run them instead of explaining them.

Read at each step

StepRead
1.agents/plugin-stack.md if present, references/data.md
1, 3references/listing.md
2, 4references/tools.md
5references/simulated-pick.md

Steps

  1. Fix the scope. The plugin, 5-10 target prompts and 3-5 competitors. Done when each is named and at least half the prompts aren't drafted, or the report says coverage is circular.
  2. Can it be reached at all? Done when the listing, the server and the target country are each marked, with checked_at printed beside any stored status. A failed mark is fix #1, and every later step is labelled conditional on fixing reachability.
  3. Audit the listing's words. Done when every target prompt is marked covered or missed, with the competitors that cover it.
  4. Audit the tools. Read them from mcp_toolset_get, the repo's server code, or tools/list output the user pastes. Done when every tool has a verdict on saying what it does, when to use it, and its annotations. With none of the three, say so in the report and go to step 5.
  5. Run the simulated pick. Done when each prompt has a pick, its reason and a run count, labelled as a simulation.
  6. Report. Ranked fixes. Done when every fix names a target prompt it should help and every number has its date and basis. If .agents/plugin-stack.md exists, add audit-discoverability: <today>, top fix: <fix> under Runs without asking, in your own words.

Report order

  1. Any Degraded: lines, then This simulation is not informative for this set if a control failed.
  2. Baseline: the best chart place from the plugin's listings_search row, with meta.as_of, so the next run has something to compare.
  3. Fixes, ranked: reachability, tool text that risks store review, prompts the plugin won and should have lost, then missed prompts, sourced ones before drafted ones. A tool fix gives the replacement text. A listing fix names the missing words and the field, and leaves the copy to optimize-listing. With no listing yet, name submit-plugin.
  4. Evidence: the coverage line, the tool verdicts, the picks with their label, the real-assistant line and what the simulation can't tell you.

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.

Check your setup, free