The Agent Skills Directory

[METADATA_POISONING]: There is a naming discrepancy between the skill metadata, which attributes authorship to "Ihor Orlovskyi," and the provided author context of "sentimony."
[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell commands (rg, git log) to inventory and audit project files for specific linguistic patterns.
[PERSISTENCE]: The skill facilitates behavioral persistence by providing a git hook (scripts/commit-msg) to enforce writing style on new commits and recommending the addition of instructions to persistent project files (CLAUDE.md or AGENTS.md) to maintain constraints across sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could contain malicious directives.
Ingestion points: Scans the local working tree and git history commit messages (SKILL.md).
Boundary markers: The skill contains explicit "Security Model" instructions telling the agent to treat scanned content strictly as data and never to follow directives found in scanned text.
Capability inventory: Includes the ability to write to files in "Fix mode" and execute local shell commands for auditing.
Sanitization: Relies on the agent's adherence to the instruction-boundary guidelines provided in the Security Model section.

negafix