knowledge-mcp: Skill for Claude Code

.claude/skills/verification-evidence/SKILL.md

verification-evidence is a skill for Claude Code from fulminate-io/knowledge-mcp. It costs 60 tokens per session (1,357 once invoked), scanned A, original, Apache-2.0.

A rulebook for treating completed implementation work as verified only after the exact checks have run and their output has been read. It covers tests, persisted edits, missing requirements, and live-system checks.

In plain words
What is it for?
Use it to verify requirements against the diff, run the same probe before and after a fix, and expose behavior that still lacks a failing check.
Why use it?
It prevents unfinished work, tests that cannot fail, and confident claims based only on existing green tests.

Skill for Claude Code

Written for Claude Code: user-invocable in frontmatter.

This is fulminate-io/knowledge-mcp's own configuration. It tells Claude Code how to work on knowledge-mcp itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything knowledge-mcp configures →

Reuse

Borrowing it

Nothing to install: this file belongs to fulminate-io/knowledge-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/fulminate-io/knowledge-mcp/main/.claude/skills/verification-evidence/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/fulminate-io/knowledge-mcp

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for verification-evidence

README.md
[![agentmods](https://agentmods.dev/badge/skills/fulminate-io/knowledge-mcp/verification-evidence/github.svg)](https://agentmods.dev/skills/fulminate-io/knowledge-mcp/verification-evidence)
Your own site
<a href="https://agentmods.dev/skills/fulminate-io/knowledge-mcp/verification-evidence"><img src="https://agentmods.dev/badge/skills/fulminate-io/knowledge-mcp/verification-evidence/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for verification-evidence

Your own site · 80×15
<a href="https://agentmods.dev/skills/fulminate-io/knowledge-mcp/verification-evidence"><img src="https://agentmods.dev/badge/skills/fulminate-io/knowledge-mcp/verification-evidence.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 60 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,357 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00060 $0.01357
Opus 5 $0.00030 $0.00678
Sonnet 5 $0.00012 $0.00271
Haiku 4.5 $0.00006 $0.00136

Measured 4d ago against content hash a883a47e554a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

verification-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 4d ago.

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.

.claude/skills/verification-evidence/SKILL.md · 113 lines

How it starts

The opening of the file, as written. The whole thing — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.

VERIFICATION-EVIDENCE — what "done" and "green" are allowed to mean

Done means verified

Mark a step complete ONLY against verification run THIS turn whose output you read. Before every completion: (1) the edit PERSISTED on disk (git status/diff or re-Read), not merely "I issued an Edit"; (2) every criterion command EXECUTED and returned this turn — real exit status, real output. If you catch a phantom completion of your own, reopen the step, redo it, disclose plainly.

Every behavior the step AND ticket specify is in the diff — not just what existing tests cover. Read step + ticket text as a CHECKLIST; decompose compounds (the second clause of one sentence is the canonical silent drop). A specified behavior with NO test that goes red when absent: ADD the failing-when-absent test, or surface the unverified requirement LOUDLY. Green proves what you BUILT works, never that you built everything specified.

Same probe, red and green

Red-first means the EXACT command and selector that will later report green was run against the unfixed tree and observed to FAIL — not a different invocation, not an inspection. Paste both raw outputs. Treat as RED-NEVER: a runner reporting no tests ran, a skipped harness, a build no-op'd by a missing tag, a selector matching nothing. If the repro PASSES against unfixed code, STOP — do not fix, do not weaken the test: either it doesn't exercise the defect, the defect isn't present (the plan's premise is wrong — a genuine finding), or it's conditional on state the test didn't establish. Determine which and report.

Never fake green

Do NOT delete, skip, or comment out tests to pass. A test failing because your step intentionally changes behavior gets updated to assert the NEW behavior — say so. Deleting a test is correct ONLY when its surface genuinely no longer exists, stated with the file:line proving it gone. Forbidden ways to leave a failure standing: "pre-existing / not my regression", "out of scope", "flaky" without a proven cited mechanism, skip/xfail/commented-out assertions. The one escalation: the failure reveals the CODE is wrong — fix it, or surface the found bug with evidence.

Read the full file on GitHub · 113 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. 4d ago First seen · 113 lines · 60 tokens per session scan A a883a47e554a

Subscribe to this mod's changes

verification-evidence is a skill published in the GitHub repository fulminate-io/knowledge-mcp (1 stars, last pushed 3d ago), licensed Apache-2.0. It adds 60 tokens to every session and 1,357 once invoked, about $0.0003 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 2026-09-04.