The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to execute discovery commands such as ls -la, tree, git status, and git log --oneline -5. These are used to collect directory structures, file metadata, and recent commit history to populate the project context summary.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill reads content from multiple local project files, including 01-requirements.md, 02-system-design.md, 04-dev-tasks.md, package.json, and Makefile (via the onboard workflow).
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" headers when interpolating data from these external files into the generated 00-context.md file.
Capability inventory: The skill possesses the Write tool for creating the summary file and Bash for local discovery and information gathering.
Sanitization: The skill includes an explicit constraint: "Sensitive information (keys, passwords) must not be included". However, there are no specific technical filters described to handle potentially malicious instructions embedded within the source documentation.

onboard