How do I install the Fpf Evidence skill?

Run npx skillmds@latest add m0n0x41d/fpf-evidence in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Fpf Evidence safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Fpf Evidence?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Fpf Evidence free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Fpf Evidence?

m0n0x41d (@m0n0x41d) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Fpf Evidence by m0n0x41d — Agent Skill

Name: Fpf Evidence
Author: m0n0x41d

File contents

What this skill IS

This is testing against reality, not documentation. You are running an experiment — defining predictions, executing tests, interpreting results. The evidence record is the lab notebook, not a post-hoc report.

Work through the template as the test procedure. Define predictions BEFORE running tests. Record what actually happened. Interpret honestly.

Feedback loop

If evidence refutes the prime hypothesis or reveals new constraints:

Update the PROB-* card (feed back to problem factory)
Consider: does the problem need reframing?
Create new ANOM-* if evidence reveals unexpected behavior

If evidence corroborates:

Note the scope limitations in G (ClaimScope)
Set valid_until (REQUIRED — expiry date or "perpetual" with justification)

Output

.fpf/evidence/EVID-${CLAUDE_SESSION_ID}--<slug>.md

Constraints (quality bar)

C1: Claim under test stated in one sentence
C2: Predictions derived BEFORE testing — what should be true, what would falsify
C3: Harness is the smallest credible check (test, typecheck, benchmark, log)
C4: Commands and outputs recorded verbatim
C5: Result labeled: corroborated | refuted | inconclusive
C6: F-G-R filled: F (formality), G (ClaimScope — set-valued, enumerate specific contexts), R (reliability 0-1)
C7: valid_until REQUIRED — set expiry date (YYYY-MM-DD) for time/version-dependent claims, or "perpetual" with justification for universal truths. Stop gate blocks if missing.
C8: If refuted: MUST update PROB-* or create new ANOM-* (feedback to problem factory). Stop gate blocks session end if refuted evidence exists without problem reframing.

Format

# Evidence Record
- **ID:** EVID-...  **Claim:** ...  **Created:** YYYY-MM-DD
- **F:** informal|structured|formalizable  **G:** {contexts}  **R:** 0-1
- **DesignRunTag:** design-time|run-time

## Predictions (defined before testing)
- If true: ...
- Would falsify: ...

## Harness
- Type: (test|benchmark|log|typecheck|manual)
- Lane: TA (theory assessment) | LA (lab assessment) | VA (verification/production assessment)
- Rationale: why this is the smallest credible check

## Commands + outputs
\`\`\`
(exact commands and raw output)
\`\`\`

## Interpretation
- Result: corroborated|refuted|inconclusive
- valid_until: ...
- Feedback to problem factory: (if refuted, what changes in PROB-*?)

---name: fpf-evidencedescription: Test a claim against reality by running the evidence procedure. Produce EVID-* record with commands, outputs, and interpretation. MUST invoke before claiming "works" or "verified".---## What this skill ISThis is **testing against reality**, not documentation. You are running an experiment — defining predictions, executing tests, interpreting results. The evidence record is the lab notebook, not a post-hoc report.**Work through the template as the test procedure.** Define predictions BEFORE running tests. Record what actually happened. Interpret honestly.## Feedback loopIf evidence **refutes** the prime hypothesis or reveals new constraints:- Update the PROB-\* card (feed back to problem factory)- Consider: does the problem need reframing?- Create new ANOM-\* if evidence reveals unexpected behaviorIf evidence **corroborates**:- Note the scope limitations in G (ClaimScope)- Set valid_until (REQUIRED — expiry date or "perpetual" with justification)## Output`.fpf/evidence/EVID-${CLAUDE_SESSION_ID}--<slug>.md`## Constraints (quality bar)- **C1:** Claim under test stated in one sentence- **C2:** Predictions derived BEFORE testing — what should be true, what would falsify- **C3:** Harness is the smallest credible check (test, typecheck, benchmark, log)- **C4:** Commands and outputs recorded verbatim- **C5:** Result labeled: corroborated | refuted | inconclusive- **C6:** F-G-R filled: F (formality), G (ClaimScope — set-valued, enumerate specific contexts), R (reliability 0-1)- **C7:** valid_until REQUIRED — set expiry date (YYYY-MM-DD) for time/version-dependent claims, or "perpetual" with justification for universal truths. Stop gate blocks if missing.- **C8:** If refuted: MUST update PROB-\* or create new ANOM-\* (feedback to problem factory). Stop gate blocks session end if refuted evidence exists without problem reframing.## Format```markdown# Evidence Record- **ID:** EVID-... **Claim:** ... **Created:** YYYY-MM-DD- **F:** informal|structured|formalizable **G:** {contexts} **R:** 0-1- **DesignRunTag:** design-time|run-time## Predictions (defined before testing)- If true: ...- Would falsify: ...## Harness- Type: (test|benchmark|log|typecheck|manual)- Lane: TA (theory assessment) | LA (lab assessment) | VA (verification/production assessment)- Rationale: why this is the smallest credible check## Commands + outputs\`\`\`(exact commands and raw output)\`\`\`## Interpretation- Result: corroborated|refuted|inconclusive- valid_until: ...- Feedback to problem factory: (if refuted, what changes in PROB-*?)```

Frequently asked questions

Fpf Evidence