Caxson

Caxson/proof-of-work

Plugin Claude Code

Engineering-discipline layer for Claude Code — evidence-first doctrine and verification skills that make the agent prove its work.

Caxson/proof-of-work

Plugin Claude Code

An engineering-discipline layer for Claude Code: an evidence-first doctrine injected every session, plus verified-delivery and evidence-first-debugging skills. Makes the agent prove its work instead of claiming success.

Caxson/proof-of-work

Instructions file CodexOpenCode

Instructions for Caxson/proof-of-work, a project described as: A disciplined-engineer + persistent-self layer for Claude Code.

Caxson/proof-of-work

Command

Evidence-first diagnosis of a bug/failure ($ARGUMENTS).

Caxson/proof-of-work

Command

Re-surface the proof-of-work Engineering Doctrine and treat it as binding for the rest of this session.

Caxson/proof-of-work

Command

Run the verified-delivery workflow on the current task (or $ARGUMENTS).

Caxson/proof-of-work

Hook

Runs when a session starts on startup, clear and compact, executing run-hook.sh. From Caxson/proof-of-work.

Caxson/proof-of-work

Skill Claude CodeCodex

Use when diagnosing a failure, bug, freeze, crash, or any "why doesn't this work" — gather concrete evidence (logs, real data, a reproduction) BEFORE naming a root cause, rule out your own code first, rank hypotheses by confidence with the supporting evidence attached, and fix the root cause rather than the symptom…

Caxson/proof-of-work

Skill Claude CodeCodex

Use when building, fixing, or implementing anything that can be run — decompose the work into steps, prove EACH step with real actions (Playwright screenshots, SQL/database checks, API round-trips, device screenshots) against real data, never mock data, and refuse to mark the task done until you have an evidence…