// Search jackfranklin/references (personal technical reference docs) for files relevant to the current task. Use when you need a reference guide on a library, API, or tool, or when the user invokes /references-search with search keywords as $ARGUMENTS.
Manage feedback and bug reports for a project using a local SQLite database. Use when the user asks to log, list, view, resolve, or work through feedback items or bugs.
Scaffold a new Deno 2 + Hono + Deno KV + Eta + HTMX app with password auth and PWA support. Use when the user wants to create a new personal web app following this stack.
Reviews code changes for hacky patterns, redundant state, parameter sprawl, and leaky abstractions. Use when the user asks to "review code quality", "check for hacky patterns", "quality review", or wants a quality-focused code review.
Reviews code changes to identify opportunities for reusing existing utilities and helpers. Use when the user asks to "review for reuse", "check for duplicate code", "find existing utilities", or wants a reuse-focused code review.
Reviews code changes for efficiency issues, unnecessary work, missed concurrency, and memory leaks. Use when the user asks to "review for efficiency", "check for performance issues", "find memory leaks", or wants an efficiency-focused code review.
Scaffold a new Vite + LitElement + IDB PWA app hosted on Netlify with a private GitHub repo. Use when the user wants to create a new personal web app following this stack.
| name | references-search |
| description | Search jackfranklin/references (personal technical reference docs) for files relevant to the current task. Use when you need a reference guide on a library, API, or tool, or when the user invokes /references-search with search keywords as $ARGUMENTS. |
Search the personal technical reference library at jackfranklin/references and return the content of matching files.
The repo always lives at $HOME/.claude/repos/references. Use the literal expanded path in every Bash command (e.g. /root/.claude/repos/references in the remote environment, or ~/.claude/repos/references locally) since shell variables don't persist between tool calls. Expand $HOME before running any command.
First check if it exists:
ls $HOME/.claude/repos/references/.git 2>/dev/null && echo "exists" || echo "missing"
If missing: clone it fresh:
mkdir -p $HOME/.claude/repos
gh repo clone jackfranklin/references $HOME/.claude/repos/references
If it exists: fetch from origin and compare local HEAD to upstream to decide if a pull is needed:
git -C $HOME/.claude/repos/references fetch origin main --quiet
git -C $HOME/.claude/repos/references rev-parse HEAD
git -C $HOME/.claude/repos/references rev-parse origin/main
If the two hashes differ, the local branch is behind — pull it:
git -C $HOME/.claude/repos/references pull --ff-only origin main
If the hashes match, the repo is already up to date — skip the pull.
The user provided: $ARGUMENTS
Split into individual search terms. If $ARGUMENTS is empty, infer relevant keywords from the current task or conversation context (e.g. the library, API, or technology being discussed).
Run two searches and deduplicate the results.
Search A — filename match:
ls $HOME/.claude/repos/references/*.md 2>/dev/null
A file matches if any search term appears in its name after treating hyphens as word separators. For example, "speech" matches browser-speech-to-text.md; "genai" matches google-genai.md.
Search B — content grep (case-insensitive, filenames only):
For each search term:
grep -ril "TERM" $HOME/.claude/repos/references/
Combine results from A and B, removing duplicates. Keep only .md files.
If more than 4 files matched, prefer files that matched on both filename and content, then filename-only, then content-only. Read the top 4.
Use the Read tool to read each matched file in full. Reference files are intentionally small and self-contained — always read the complete file, never truncate.
If Step 4 returned nothing, list all available reference files:
ls $HOME/.claude/repos/references/*.md 2>/dev/null
Display as a bullet list (strip the path prefix and .md suffix for readability) and ask the user which one to read, or pick the most likely candidate based on context.
Explain which file(s) you found and why they're relevant, then present the full reference content in a way that directly addresses the current task. Do not summarize or truncate — complete, accurate reference material is the deliverable.
If multiple files matched, separate their contents with a heading per file.