The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution using tools like ls, grep, find, cat, mkdir, ln, git submodule, and jq to map repository topology and extract API contracts across monorepos and multi-repo environments.
[COMMAND_EXECUTION]: It employs relative path traversal (cat ../*/.git/config) to access and analyze sibling directories, which may be outside the scope of the primary repository the agent is currently operating within.
[CREDENTIALS_UNSAFE]: By reading .git/config files with the command cat ../*/.git/config 2>/dev/null | grep "url", the skill risks exposing sensitive authentication tokens or credentials that might be stored within Git remote URLs.
[PROMPT_INJECTION]: A significant indirect prompt injection surface exists because the skill processes untrusted data from various files throughout the workspace to create summary artifacts.
Ingestion points: Processes content from package.json, source code (.ts, .py), and schema files across the entire discovered workspace.
Boundary markers: No explicit delimiters are used to wrap ingested content or warn the agent about potential malicious instructions within analyzed data.
Capability inventory: Utilizes shell execution (grep, cat, find) and file system write access to generate and update documentation artifacts.
Sanitization: There is no evidence of content sanitization or validation before interpolation into generated summary files.
[COMMAND_EXECUTION]: The skill encourages the establishment of persistent development workflows by suggesting the creation of Git pre-commit hooks and GitHub Actions, which execute code automatically during the development lifecycle.

workspace