# Evidence Recording

> Records the real, tool-produced output of each required check as Evidence — never a summarized or claimed result (docs/PROCESS_CONTRACT.md, stage TESTING; docs/ENGINEERING_STANDARDS.md §9).

- Skill: `erik-corder/evidence-recording` (Agent Skill, multi-file: 7 files)
- Install (CLI): `npx skillmds@latest add erik-corder/evidence-recording`
- Raw SKILL.md: https://api.skillmd.com/api/skills/erik-corder/evidence-recording/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: erik-corder (https://skillmd.com/u/erik-corder)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/erik-corder/evidence-recording

---


# Evidence Recording

This Skill deliberately declares **no AI capability at all** — it is the
one Skill in this pack whose entire job is mechanical: take what a real
tool actually reported, and write it down. This mirrors the domain
model's own distinction between `Artifact` (a human-directed deliverable)
and `Evidence` (a deterministic, tool-produced verification result) in
`backend/src/skillbase_backend/domain/entities.py`.

## Why this is its own Skill, separate from `verification`

`verification` (in the `engineering` pack) is the Skill that actually
*runs* the required checks. This Skill is what turns each check's real
result into a durable, attributable `Evidence` record. Splitting them
keeps "did the check run for real" (verification) and "is the result
properly recorded" (evidence-recording) independently reviewable.