document-with-evidence

A format and process for recording curated knowledge in Markdown files with evidence and source details. Each entry has a stable identifier and can include when and where the evidence was captured.

In plain words
What is it for?
Use it to record a research finding, technical decision, measured result, or other knowledge item that should be checked again later.
Why use it?
It makes important findings, decisions, and measured facts easier to verify and update. New corrections can replace older entries without reusing identifiers.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/everywan-dev/claude-code-engineering/document-with-evidence
Any agent
npx skills add everywan-dev/claude-code-engineering --skill document-with-evidence
Clone the repo
git clone --depth 1 https://github.com/everywan-dev/claude-code-engineering

Made for: Claude Code, Codex.

Per session 92 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 871 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00092 $0.00871
Opus 5 $0.00046 $0.00436
Sonnet 5 $0.00018 $0.00174
Haiku 4.5 $0.00009 $0.00087

Measured yesterday against content hash b5f4858293e1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

document-with-evidence scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/document-with-evidence/SKILL.md · 70 lines

How it starts

The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Document with evidence

A curated-knowledge unit is one Markdown file under knowledge/, with frontmatter carrying the base contract. Write the file directly (or via your usual editing tools) -- there is no validated_memory subcommand that creates one; validate only checks what you wrote.

The base contract, field by field

---
id: kb-0001                    # required; stable, unique; matches the adopter's id_prefix
evidence: measured              # required; measured | verifiable | hypothesis
supersedes: []                  # optional; ids this unit supersedes (see supersede-outdated-docs)
anchors:                        # optional; without anchors a unit cannot expire
  - system: <system-name>
    kind: <probe-kind>          # must match a kind registered in validated-memory.md
    captured_at: 2026-08-12T00:00:00Z
    payload: {}                 # interpreted by the probe, not by validate
provenance: []                  # optional; where the native artifact lives
---
  • id -- check id_prefix in validated-memory.md for the scheme this project follows. Never reuse an id once written; a correction is a new unit (see supersede-outdated-docs).

  • evidence -- pick the state the knowledge actually has right now, not the state you hope to reach (canonical definitions: the README's "Base contract" section):

    • measured -- directly observed or computed, with a way to re-check it.
    • verifiable -- not directly measured, but checkable by someone who follows the provenance.
    • hypothesis -- a claim not yet checked.

    Never promote a unit's evidence state by editing it in place because you have grown more confident. Promotion is not a state change on the same file; it is a new unit with better evidence that supersedes this one (supersede-outdated-docs). Evidence state moves forward only by writing new knowledge, never by conviction.

  • anchors vs provenance -- do not mix the two planes:

    • An anchor is a probeable claim: "system X, of kind K, was true as of captured_at, described by this payload" -- exactly what detect-stale-documentation re-checks later. A unit with no anchors can never be checked for freshness (a WARNING, not an ERROR).
    • Provenance is not probed: it records where the native artifact (a query, a document, a conversation) lives, for a human to go look. Put a link or a citation there, not something you expect a probe to check.

Read the full file on GitHub · 70 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 70 lines · 92 tokens per session scan A b5f4858293e1

Subscribe to this mod's changes

document-with-evidence is a skill published in the GitHub repository everywan-dev/claude-code-engineering (2 stars, last pushed 13d ago), licensed Apache-2.0. It adds 92 tokens to every session and 871 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.