The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute a variety of development-related shell commands and tools, including git, make, pnpm, npm, uv, xcodebuild, and adb. It specifically favors using make targets for tasks like testing, linting, and building if a Makefile is present in the project root.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves reading and acting upon content from files such as plan.md and spec.md, which creates an avenue for malicious instructions to be processed by the agent.\n
Ingestion points: The skill reads from docs/plan/*/plan.md, docs/plan/*/spec.md, docs/workflow.md, and CLAUDE.md as described in SKILL.md.\n
Boundary markers: There are no explicit delimiters or safety instructions provided to the agent when it ingests content from these documentation files.\n
Capability inventory: The skill has extensive capabilities including shell execution (Bash), file modification (Write, Edit), and environment management through package managers.\n
Sanitization: The skill lacks sanitization for instructions or commands extracted from the implementation plans and specifications; it specifically instructs the agent to run commands found within spec.md acceptance criteria.\n- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes load-time command execution placeholders to populate context information.\n
Evidence: SKILL.md contains the patterns !git branch --show-current, !git status --short, and !git log --oneline -3 to fetch repository metadata.\n- [EXTERNAL_DOWNLOADS]: The skill automates the installation of development dependencies and git hooks using standard package management tools.\n
Evidence: It executes commands such as pnpm prepare, uv run pre-commit install, and lefthook install to ensure the environment is configured correctly.

solo-build