The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs aggressive directive language designed to override agent autonomy and user intent. Phrases like 'No exceptions, no rationalizations, no shortcuts' and 'applies to EVERY session' attempt to force the agent into a specific behavior pattern regardless of the complexity or nature of the user's request, specifically targeting and instructing the agent to ignore user rationalizations for skipping the defined workflow.
[PERSISTENCE]: The skill mandates a 'sidecar' outcome emission where it writes JSON metadata to ~/.claude/skill-analytics/last-outcome-workflow-enforcer.json at the conclusion of every session. This creates a persistent trail of interaction data in a system-level configuration directory.
[DATA_EXFILTRATION]: The automated logging mechanism tracks session IDs, timestamps, and specific metrics about 'violations caught' and agents recommended. While framed as internal analytics, this metadata exposure creates a detailed log of user activity and project workflow patterns that is automatically generated and stored on the local filesystem without per-session user consent.

workflow-enforcer-skill