The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Node.js utility scripts/artifact.mjs which performs file system operations and calls the git binary using child_process.execFile. The arguments passed to git are static strings (e.g., rev-parse, branch) used to determine repository state.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data and is explicitly instructed to follow instructions contained within that data.
Ingestion points: SKILL.md instructs the agent to read all artifacts in a task directory and the source code referenced by research files.
Boundary markers: There are no explicit instructions or delimiters defined to help the agent distinguish between data and instructions when reading these files.
Capability inventory: The agent has access to file manipulation tools (Read, Write), search tools (Grep, Glob), and shell execution capabilities (Bash, and the artifact.mjs script).
Sanitization: scripts/artifact.mjs validates metadata attributes to prevent corruption of the artifact's Markdown structure, but it does not sanitize or filter the content of the implementation plan body against embedded prompts.

create-plan