From vibe-sentinel
Use before implementation, bug fixes, refactors, parser changes, Playwright changes, CI changes, or architecture changes to identify, verify, and record assumptions before editing code.
How this skill is triggered — by the user, by Claude, or both
Slash command
/vibe-sentinel:assumption-killer Describe the implementation task whose assumptions must be gatedDescribe the implementation task whose assumptions must be gatedThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill to prevent implementation based on confident but unverified assumptions. It creates an Assumption Gate that records what the agent believes, how each belief was checked, and whether implementation is allowed.
Use this skill to prevent implementation based on confident but unverified assumptions. It creates an Assumption Gate that records what the agent believes, how each belief was checked, and whether implementation is allowed.
Use before modifying production code, tests, config, parser logic, Playwright flows, build scripts, CI, release logic, schemas, data models, API contracts, security rules, permissions, or architecture.
Do not use for pure read-only explanation, tiny formatting edits, typo fixes in documentation, or tasks where the user explicitly asks only for analysis and no implementation.
Create or update these files before implementation:
.agent/session/assumption-gate.json.agent/session/assumptions.mdInitialize them with:
node plugins/assumption-killer/bin/assumption-gate.mjs init
Use only these categories: repo_structure, existing_behavior, test_behavior, runtime_behavior, tooling_behavior, parser_behavior, ux_behavior, data_model, api_contract, security_or_permissions, backwards_compatibility, deployment_or_ci.
low: Wrong assumption causes minor rework or local confusion.medium: Wrong assumption can break a narrow workflow or test.high: Wrong assumption can produce incorrect implementation, regressions, or unsafe behavior.critical: Wrong assumption can cause data loss, security exposure, destructive action, production impact, or broad architectural damage..agent/session/assumption-gate.json and summarize it in .agent/session/assumptions.md.high or critical assumption has status unknown.disproven, update the plan before implementation.node plugins/assumption-killer/bin/assumption-gate.mjs check.frontend/package.json defines test: vitest run."pattern.call_expression because backend/src/indexing/parser.rs maps that capture name."symbolId because frontend/src/api/types.ts exports it and current tests assert it."Weak assumptions are not evidence. Replace them with checked files, commands, or contracts.
Evidence must be concrete and reproducible. Prefer exact file paths, command names, schema paths, test names, or API contracts. Do not cite memory, intuition, naming guesses, or old run results as verified evidence unless you also checked current repository state.
Each verified assumption must include non-empty evidence, and should include files_checked or commands_run when applicable.
high or critical assumption is unknown.disproven assumption without updating the implementation decision.verified without evidence.check command.Report whether the Assumption Gate passed. Mention any remaining medium or low unknowns as open risk. If the check failed, state that implementation is blocked and name the blocking assumptions.
node plugins/assumption-killer/bin/assumption-gate.mjs init
node plugins/assumption-killer/bin/assumption-gate.mjs check
node plugins/assumption-killer/bin/assumption-gate.mjs summary
Run from the target repository root so artifacts are created under .agent/session/.
Codex: use this skill before edits and run check before the final response.
GitHub Copilot: paste AGENTS.fragment.md into repository instructions or wire the example hook where supported.
Claude: paste custom-instructions.fragment.md into custom instructions or install this plugin where supported.
Hooks can enforce artifact validity, but they cannot force the agent to choose this skill. Keep the instructions and hooks together.
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub sammroberts/agentic-engineering --plugin vibe-sentinel