The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/graveyard.py executes local git commands (e.g., git log, git ls-files) via subprocess.run to extract commit history and metadata. These calls use structured argument lists and do not invoke a shell, preventing command injection vulnerabilities.
[DATA_EXPOSURE]: The skill performs read-only scanning of local project directories and git metadata. It is designed for offline execution, contains no network-related code, and includes a --redact feature to anonymize project names in reports.
[PROMPT_INJECTION]: The skill processes untrusted data from local git histories, creating a surface for indirect prompt injection.
Ingestion points: Commit messages and filenames are extracted by scripts/graveyard.py and provided to the agent.
Boundary markers: SKILL.md defines a structured report format but lacks explicit delimiters or instructions to ignore embedded commands within the raw git output.
Capability inventory: The skill's capabilities are limited to local filesystem traversal, reading git metadata, and writing to a local state file.
Sanitization: No specific sanitization or filtering is applied to commit messages before they are processed by the agent.

project-graveyard