The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses automated hooks to run local shell commands and scripts. The PreToolUse hook executes cat task_plan.md to refresh context, and the Stop hook runs check-complete.sh to verify task completion. These commands are localized to the project environment and support the skill's operational logic.
[PROMPT_INJECTION]: The skill automatically reads content from task_plan.md into the agent's context during tool usage.
Ingestion points: The PreToolUse hook in SKILL.md reads the first 30 lines of task_plan.md before executing Write, Edit, or Bash tools.
Boundary markers: Content is injected without specific delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has access to Bash, Write, Edit, and WebFetch as defined in the YAML frontmatter.
Sanitization: There is no validation or sanitization of the file content before it is read into the context. This creates a surface for indirect instructions if the task plan file is modified by an external untrusted source.

planning-with-files