← ClaudeAtlas

project-prospectorlisted

Discover, catalog, and rank everything a person has built or sketched on a machine by fanning out parallel read-only investigation agents over the filesystem, then synthesizing a tiered ranking by idea-novelty and leverage. The defining signal is a whole-machine survey of the user's OWN work — every project, repo, half-built thing, abandoned experiment, or loose idea (notes, downloads, transcripts included), usually wanting a ranking or verdict: most original, strongest, worth finishing. Trigger on asks like "what have I built on this laptop", "take stock of my half-finished repos", "rank my projects by which ideas are most original", or "fan out agents and catalog my side projects" — even without the words "agents" or "skill". Supports a time window and scope exclusions (e.g. "except my job's repo"). Strictly read-only. Do NOT use it for work inside a single repo (its files, PRs, TODOs), bare repo listing, choosing the next feature, or surveying projects that aren't the user's.
88plug/project-prospector · ★ 0 · AI & Automation · score 60
Install: claude install-skill 88plug/project-prospector
# Project Prospector Surface and rank a person's body of work on a machine. The output is a **tiered ranking of projects/ideas/codebases** — strongest (most original ideas, most leverage) first — built from a parallel sweep that no single linear pass would find. The name is the intent: you're prospecting a messy filesystem for the few genius ideas buried in it. ## Operating context: run this from the main conversation This skill *spawns* investigation agents, so it belongs in the main agent loop where the Agent/Task tool is available. If you find yourself running it somewhere that can't spawn sub-agents (e.g. you are yourself a sub-agent, and nested agents are disallowed), don't abort — **fall back to running the same two-pass structure inline and sequentially yourself.** Use the identical scout commands, per-cluster git-log/mtime sweep, and the five Pass-2 blind-spot checks; you only lose parallelism, not coverage. Say so in one line at the top of your output so the reader knows the run was serial. One self-reference trap to close every time: **this skill and its working files live on the machine you're scanning.** Exclude the prospector skill directory and any `*-workspace/` it created from the catalog — they're this tool's own scaffolding, not the user's projects — and never treat your own draft report as a finding. If you write the final report to disk and the Write tool is blocked for it, fall back to a Bash heredoc (`cat <<'EOF' > report.md`). **The machine is a mo