The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted text from user files or git staged changes. Because the content of these files is processed to generate linguistic reports without strict sanitization or explicit boundary instructions for the agent, it represents a potential surface for indirect prompt injection if a file contains malicious instructions formatted as prose.
Ingestion points: git diff output and file content read from disk.
Boundary markers: While the output format wraps quotes around findings, there are no specific system-level instructions provided to isolate the ingested text from the agent's logic during the analysis phase.
Capability inventory: Bash, Grep, Read, Glob.
Sanitization: No explicit sanitization of the input text is mentioned or implemented in the instructions.
[PERSISTENCE]: The skill provides instructions and a conceptual bash snippet for installing a git pre-commit hook in .git/hooks/pre-commit. Although this is a standard developer integration and the skill explicitly states it will not perform the installation automatically, this mechanism allows for automatic code execution during the git lifecycle.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute git metadata commands (git diff --cached, git diff HEAD~1 HEAD, etc.) and a local Python script (python3 writer/scripts/lint.py). These commands are used for scanning and reporting, which is the core functionality, but they involve executing logic based on the repository state.

style-check