The Agent Skills Directory

[SAFE]: The skill implements proactive security measures by explicitly instructing the agent to never read or reference sensitive files such as .env variants, API keys, tokens, or private keys.
[COMMAND_EXECUTION]: Uses the Execute tool to run standard git commands (git diff, git log) for the purpose of identifying recent changes and risks. This behavior is restricted to the repository's metadata and is aligned with the skill's stated purpose.
[EXTERNAL_DOWNLOADS]: Includes a reference to the vendor's documentation site (docs.factory.ai). This is a legitimate resource from the skill's author and does not present a security risk.
[PROMPT_INJECTION]: The skill uses a 'solemn chaplain' persona to guide the stylistic tone of its output. This role-play is benign and does not attempt to override the underlying safety protocols of the agent.
[PROMPT_INJECTION]: Analysis of indirect prompt injection vulnerability surfaces:
Ingestion points: Git commit history (git log), code diffs (git diff), and file contents via Grep (SKILL.md).
Boundary markers: Absent; there are no specific delimiters defined to separate untrusted code data from the agent's instructions.
Capability inventory: The skill can execute shell commands via Execute, read files, and interact with the user via AskUser.
Sanitization: No explicit sanitization or validation of the ingested repository data is performed before processing.

ship-ritual