This skill provides a communication policy for agents: verify claims, prioritise facts over validation, and deliver direct, evidence-backed assessments. It codifies phrases to avoid, recommended phrasing, and a decision flow for investigating user claims. Practically, the skill helps an agent respond to technical questions, debugging reports, or design proposals with rigorous checks, respectful disagreement, and clear reasoning.
Use when a user makes technical claims, asks for evaluations, or proposes solutions where correctness matters (debugging, code review, incident triage, technical design). Activate when evidence is required or when the agent should avoid appeasing the user at the cost of accuracy.
Best-suited for assistant/Claude-style agents and code-review assistants that can access code, logs, or documents to verify claims. It complements agents that have read/write capabilities and access to diagnostic tools.
Pure-guideline skill with no scripts or executable code. Well-structured SKILL.md providing behavioral rules for agents to prioritise accuracy over agreeableness. Clear do/don't examples and scenario patterns. No security concerns whatsoever — no network calls, no credentials, no destructive commands.
Clean, well-written guideline skill. No executable surface to test. Frontmatter is complete with name, user-invocable flag, description, and allowed-tools.