discover-agentlisted

# Discover agent Query: `$ARGUMENTS` — a keyword, domain, or task description. ## Render-time preflight (inline dynamic context) These checks run BEFORE Claude reads the workflow, so the model sees the preflight state with zero extra tool turns. Each `` !`cmd` `` line is replaced by command output at render time. - FTS5 index present: !`test -f ~/.claude/cache/maxvision/index.db && echo "INDEX_OK" || echo "INDEX_MISSING"` - Cheatsheet path: !`for p in .claude/CLAUDE-agents.md "$HOME/.claude/CLAUDE-agents.md"; do [ -f "$p" ] && echo "$p" && break; done; true` ## Workflow ### 1. Honor preflight result If the render-time line above printed `INDEX_MISSING`, abort immediately with the structured response below and propose `/maxvision:index-catalog`: ```json {"matches":[],"error":"index missing","suggestion":"Run /maxvision:index-catalog first"} ``` Otherwise the index file exists; proceed. ### 2. Resolve cheatsheet for trusted-mappings The render-time `Cheatsheet path:` line above resolved the first existing of `./.claude/CLAUDE-agents.md` (project) or `~/.claude/CLAUDE-agents.md` (user). When non-empty, that file is the authoritative trusted-mappings source. Agents listed there get a `★ in cheatsheet` badge in the output. ### 3. Build BM25 SQL with column weights and `kind='agent'` filter Same pattern as `query-index` — heredoc + SQL-literal escape (NEVER raw `$ARGUMENTS` in SQL). Column weights identical: `name(1.5) > description(1.0) > when_to_use(1.0) > body(0.8