The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions explicitly direct the agent to include "credentials ... referenced" in the Markdown handoff file under the 'Context' section. Saving raw credentials to a shared network drive in plain-text markdown format is an unsafe practice that exposes secrets to anyone with access to the shared mount.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8). It collects untrusted data and incorporates it into a file intended to serve as instructions for another AI agent ('The reader is Claude Code on a different machine').
Ingestion points: Data is ingested from git log, git status, and the current conversation context (SKILL.md).
Boundary markers: The skill does not define delimiters or provide 'ignore embedded instructions' warnings when interpolating untrusted data into the handoff template.
Capability inventory: The skill uses the Write tool to save files and the Bash tool to execute commands (SKILL.md).
Sanitization: No sanitization or escaping is performed on the gathered context before it is written to the output file.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute standard system discovery commands including hostname, pwd, date, and git (SKILL.md).

save