The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/init_vault.py uses subprocess.run to execute standard version control commands (git init, git add, git commit, git push) and the GitHub CLI (gh repo view). These commands are necessary for the skill's primary function of initializing and syncing a private repository. The script correctly implements a privacy check to ensure user data remains confidential.
[PROMPT_INJECTION]: In Phase 3, the skill processes unstructured user input to populate context files like business-profile.md and watchlists.md. This represents an indirect prompt injection surface because the agent stores this untrusted data and may later process it as instructions. * Ingestion points: User-provided context dump during Phase 3 onboarding. * Boundary markers: No markers are used when writing the data to markdown files. * Capability inventory: The skill has the ability to write files and execute git commands. * Sanitization: The skill does not perform validation or sanitization on the user-provided data before writing it.

start-second-brain