mode-04-abstention

SKILLWorkflowcommunity
v0.0.0berto-playUnknownUpdated 2mo agoSource →

A session-level enforcement MODE that kills answering when the model should say "I don't know": fabricating under uncertainty, overclaiming confidence, guessing on unknowables, treating post-cutoff events as current, and false precision. Before asserting a claim it checks the claim is grounded, know

Community-submitted skill. Not yet reviewed by the Forge team. Full prompt content may not be available.Request review →
1Clients
1Formats
2mo agoLast update
Skill
Authorberto-play
Version0.0.0
LicenseUnknown
CategoryWorkflow
Formatsskill.md
PromptNot published
Compatibility
Claude✓ Supported
Cursor—
Copilot—
ChatGPT—
Gemini—
About

A session-level enforcement MODE that kills answering when the model should say "I don't know": fabricating under uncertainty, overclaiming confidence, guessing on unknowables, treating post-cutoff events as current, and false precision. Before asserting a claim it checks the claim is grounded, known, verifiable, and in scope; when grounding fails it abstains or caveats instead of guessing. Runs A

Keywords
skillclaude

No dependency coverage

This entry publishes no npm package, so Forge has no dependency tree for it. That is a gap in coverage — not a statement that it has no dependencies.