The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute multiple shell commands using git -C to manage the lifecycle of idea captures. This includes git pull, git add, git commit, and git push. The instructions explicitly authorize the agent to bypass standard safety confirmations for these operations within the context of the <ideas-repo>. While functional, this represents a high level of autonomy for network and file system operations.
[INDIRECT_PROMPT_INJECTION]: The skill serves as a conduit for untrusted data to be placed in persistent, shared storage.
Ingestion points: The skill captures all text following the /save-idea command as the entry (file SKILL.md).
Boundary markers: The instructions do not specify any delimiters (like XML tags or triple quotes) to wrap user-provided content when it is written to the markdown files, nor do they include warnings for future agents to ignore embedded instructions.
Capability inventory: The agent has the capability to write to local files (VIDEO-IDEAS.md, TOPICS.md, etc.), stage changes (git add), and transmit those changes to a remote server (git push) in SKILL.md.
Sanitization: The skill specifically mandates that user wording be preserved "verbatim" or "tightened" without changing meaning, and does not perform any escaping or filtering of potentially malicious instructions or markdown syntax within the user's idea text.

save-idea