The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/cartographer.mjs executes several local binaries including git, gitnexus, ast-grep, and rg using child_process.spawn. It correctly uses the shell: false option and array-based arguments to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading source code and tool outputs from the local repository, which represents a potential indirect prompt injection surface. Ingestion points: scripts/cartographer.mjs reads output from gitnexus, ripgrep, and ast-grep which process repository files. Boundary markers: The skill instructions in SKILL.md advise the agent to cite exact path:line and record unknowns rather than guessing. Capability inventory: The skill can execute code analysis tools via child_process.spawn and write map files to the .ramos/maps/ directory. Sanitization: Tool outputs are parsed as JSON, but the raw content of code symbols is presented to the agent in markdown format.
[PERSISTENCE]: The skill includes a library scripts/_lib/doctor.mjs capable of installing lifecycle hooks into the agent's configuration file (.claude/settings.local.json). This allows the agent to execute specific scripts during session start, prompt submission, or session stop events.

cartographer