perspectives-skill
CLAUDE.md1 starsChanged 22 days agoMITView on GitHub
Public Claude Code skill: multi-persona evaluation (gather → confirm → T1 Reveal / T2 Analyze / T3 Prove). End-to-end: a fresh claude -p --session-id <uuid> in a scratch folder with a toy subject, then --resume <uuid> per turn, playing the user. Check: ≤3 gather questions, one Confirm message (understanding + mode + angles), the Unaskable, a stance, T3 tool resolution shown in the cost gate, agents that cite real evidence. Users clone into ~/.claude/skills/perspectives. Keep the repo root installable as-is.
In this repositoryOne other file it gives its agents.
- perspectivesSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
No personal data. This repo is public.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
perspectives-skill
Public Claude Code skill: multi-persona evaluation (gather → confirm → T1 Reveal / T2 Analyze / T3 Prove).
Read first — all of it, before editing
SKILL.md— the flow and every gatepersona-engine.md— casting rules (mandatory, referenced from SKILL.md)agents/*.md— T3 protocols; skim the "Your Assignment" and gate sections of eachREADME.md— the public face; must match SKILL.md
Invariants
- Main file is
SKILL.md(uppercase). Case-sensitive systems won't loadskill.md. - Tier names are Reveal / Analyze / Prove everywhere. Grep for old names (Quick Council, Deep Analysis, Advanced) after edits.
agents/*.mdare prompt templates, not registered subagents. SKILL.md's Dispatch step pastes the file + an Assignment block. A field added to an agent's "Your Assignment" needs nothing else; a new Assignment-block field must be handled in every agent that uses it.- Tool access is never hard-coded to one MCP. Ladder: already loaded → installed → temporary (
npx -y/uvx/pipx run, only if approved at the cost gate) → fallback with lowered confidence. - No personal data. This repo is public.
Testing a change
End-to-end: a fresh claude -p --session-id <uuid> in a scratch folder with a toy subject, then --resume <uuid> per turn, playing the user. Check: ≤3 gather questions, one Confirm message (understanding + mode + angles), the Unaskable, a stance, T3 tool resolution shown in the cost gate, agents that cite real evidence.
Install location
Users clone into ~/.claude/skills/perspectives. Keep the repo root installable as-is.
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
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.