agentic-watchdog

Compare current implementation against spec artifacts — requirement-by-requirement alignment check

Run an on-demand session health audit — checks for spec drift, sycophantic confirmation, agentic loops, context degradation, silent failures, and premature completion

Analyze recent tool calls for repetition patterns — detect agentic loops and suggest alternative approaches

Spec-vs-code alignment specialist for detecting implementation drift. <example> User: Check if our code still matches the spec Use: Yes — compare implementation files against spec artifacts for alignment <commentary>User wants spec compliance check. The drift-detector reads both specs and code to find divergence.</commentary> </example> <example> User: Are we still on track with the requirements? Use: Yes — verify current implementation covers all specified requirements <commentary>User wants a requirements coverage check. The drift-detector builds an alignment matrix from spec to code.</commentary> </example> <example> User: Write a new feature for user authentication Use: No — this is feature implementation, not drift detection <commentary>Route to feature development. The drift-detector only compares existing code against specs.</commentary> </example> <example> User: Something feels off with the session quality Use: No — general session health concern, route to watchdog-analyst <commentary>The drift-detector only handles spec-vs-code alignment. For broader session health, use the watchdog-analyst agent.</commentary> </example>

Deep session health analysis when something feels wrong. <example> User: Something seems off with this session, can you check? Use: Yes — perform deep analysis of conversation patterns for failure indicators <commentary>User suspects quality issues. The watchdog-analyst reviews the full session for all 6 failure patterns.</commentary> </example> <example> User: I feel like we're going in circles Use: Yes — analyze recent tool calls for loop patterns and context degradation <commentary>User detects a potential agentic loop. The watchdog-analyst checks for repeated edits, retried commands, and oscillation patterns across the full session.</commentary> </example> <example> User: Fix this bug in the authentication module Use: No — this is a debugging task, not a session quality concern <commentary>Route to systematic-debugging or feature development, not watchdog analysis.</commentary> </example> <example> User: Add a new endpoint to the API Use: No — this is a feature implementation request <commentary>Route to feature development. The watchdog-analyst only diagnoses session quality, not implementation tasks.</commentary> </example>

This skill should be used when the user asks to "check session health", "detect failure patterns", "run a health audit", "find agentic loops", "check for spec drift", "verify session quality", or needs to identify and remediate failure patterns in agentic coding sessions including spec drift, sycophantic confirmation, silent failures, tool selection errors, agentic loops, and context degradation.

4 hooks across 3 events