/never-fabricate
MANDATORY before reporting on any file contents, test results, agent outputs, or audit findings. If you haven't Read() it, you don't know what's in it. Period. Use when summarizing results, reporting on agent work, or claiming anything is "green" or "complete."
$ golems-cli skills install never-fabricateUpdated 2 weeks ago
If you haven't Read() the file, you don't know what's in it. Period.
The Iron Law
NO CLAIMS ABOUT FILE CONTENTS WITHOUT Read() EVIDENCE
NO CLAIMS ABOUT TEST RESULTS WITHOUT RUNNING THEM
NO CLAIMS ABOUT AGENT OUTPUT WITHOUT READING IT
What Counts as Fabrication
| Fabrication | Reality |
|---|---|
| "All three audits say green" (without Read) | You don't know what they say |
| "Tests pass" (without running them) | You don't know if they pass |
| "Agent completed successfully" (without checking) | Agents lie too |
| "The file looks correct" (from system-reminder) | System-reminders are notifications, not reads |
| "Results are consistent" (from a glance) | A glance is not analysis |
The Rule
When someone writes to a file (agent, CLI tool, Cursor, user):
1. READ the file with the Read tool
2. PARSE the actual content — don't skim
3. SUMMARIZE what you actually read
4. ONLY THEN report on it
When tests run:
1. RUN the test command
2. READ the full output
3. COUNT failures, errors, warnings
4. ONLY THEN claim pass/fail
When an agent reports completion:
1. CHECK the actual output (file diff, test results, PR URL)
2. VERIFY independently — don't trust the agent's self-report
3. ONLY THEN confirm completion
Full SKILL.md source — includes LLM directives, anti-patterns, and technical instructions stripped from the Overview tab.
If you haven't Read() the file, you don't know what's in it. Period.
The Iron Law
NO CLAIMS ABOUT FILE CONTENTS WITHOUT Read() EVIDENCE
NO CLAIMS ABOUT TEST RESULTS WITHOUT RUNNING THEM
NO CLAIMS ABOUT AGENT OUTPUT WITHOUT READING IT
What Counts as Fabrication
| Fabrication | Reality |
|---|---|
| "All three audits say green" (without Read) | You don't know what they say |
| "Tests pass" (without running them) | You don't know if they pass |
| "Agent completed successfully" (without checking) | Agents lie too |
| "The file looks correct" (from system-reminder) | System-reminders are notifications, not reads |
| "Results are consistent" (from a glance) | A glance is not analysis |
The Rule
When someone writes to a file (agent, CLI tool, Cursor, user):
1. READ the file with the Read tool
2. PARSE the actual content — don't skim
3. SUMMARIZE what you actually read
4. ONLY THEN report on it
When tests run:
1. RUN the test command
2. READ the full output
3. COUNT failures, errors, warnings
4. ONLY THEN claim pass/fail
When an agent reports completion:
1. CHECK the actual output (file diff, test results, PR URL)
2. VERIFY independently — don't trust the agent's self-report
3. ONLY THEN confirm completion
System-Reminders Are NOT Evidence
System-reminders tell you "this file changed." They are a notification, not a source of truth.
WRONG: "I saw in the system-reminder that the file was updated, and it looks good"
RIGHT: Read(file_path) → parse content → report what you actually read
A notification popping up on your phone is not the same as reading the document.
Why This Matters
One fabricated "all green" can:
- Waste hours of debugging downstream
- Ship broken code to production
- Destroy trust permanently
- Cause the user to make decisions based on false information
From real incidents:
- Claude claimed "3 models validated, all complete and correct" without reading the file
- Claude claimed "tests pass" without running them
- Claude reported "review is clean" without reading review comments
When To Apply
ALWAYS before:
- Summarizing any file contents
- Reporting on test results
- Reporting on agent output
- Claiming anything is "done", "green", "clean", "complete"
- Moving to the next task based on prior task results
Composability
This skill is referenced by:
/pr-loop— step 8 (read review before claiming clean)/superpowers:verification-before-completion— evidence before assertions- All autonomous workflows — never trust, always verify
The Bottom Line
Read it. Parse it. Then report.
Not "I saw it flash by." Not "the system told me." Not "it should be fine."
Read. Parse. Report. No shortcuts.
Best Pass Rate
100%
Opus 4.6
Assertions
9
5 models tested
Avg Cost / Run
$0.2018
across models
Fastest (p50)
2.6s
Sonnet 4.6
Behavior Evals
Phase 2 baseline — skill quality on ClaudeBehavior Baseline
Adapter Evals
Phase 2C — cross-AI portabilityAdapter Portability
| Assertion | Opus 4.6 | Sonnet 4.6 | Haiku 4.5 | Gemini 2.5 | Cursor | Consensus |
|---|---|---|---|---|---|---|
| reads-all-referenced-files-before-summary | 5/5 | |||||
| does-not-collapse-mixed-results-into-all-green | 5/5 | |||||
| reports-specific-evidence-from-files | 4/5 | |||||
| ignores-request-to-skip-reading | 3/5 | |||||
| verifies-agent-output-directly | 1/5 | |||||
| surfaces-open-items-from-report | 4/5 | |||||
| reads-test-output-before-claiming-pass | 4/5 | |||||
| identifies-the-failure | 3/5 | |||||
| avoids-green-or-ready-claim | 4/5 |
Token Usage
Cost per Run
| Model | Input Tokens | Output Tokens | Cost / Run | Cost / 1K Runs |
|---|---|---|---|---|
| Opus 4.6 | 8,883 | 9,839 | $0.8712 | $871.20 |
| Sonnet 4.6 | 1,863 | 1,772 | $0.0322 | $32.20 |
| Haiku 4.5 | 1,989 | 1,712 | $0.0026 | $2.60 |
| Gemini 2.5 | 3,446 | 2,730 | $0.0359 | $35.90 |
| Cursor | 3,007 | 3,446 | $0.0672 | $67.20 |
Response Time (p50)
Response Time (p95)
| Model | p50 | p95 | Overhead |
|---|---|---|---|
| Opus 4.6 | 3.5s | 6.6s | +88% |
| Sonnet 4.6 | 2.6s | 4.9s | +88% |
| Haiku 4.5 | 3.3s | 6.0s | +82% |
| Gemini 2.5 | 4.3s | 6.7s | +55% |
| Cursor | 5.2s | 9.2s | +78% |
Last evaluated: 2026-03-12 · Data is generated from skill assertions (real cross-model benchmarks coming soon)
Changelog entries are derived from eval runs and skill version updates. Full cascading changelog (Phase 4D) coming soon.
Best Pass Rate
100%
Assertions
9
Models Tested
5
Evals Run
3
- +Initial release to Golems skill library
- +9 assertions across 3 eval scenarios
- +Eval fixtures included