The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute a local Python utility, scripts/subspawn_plan.py, to generate orchestration plans and validate role templates.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to orchestrate subagents by generating prompts based on user-supplied tasks, which creates a potential surface for indirect prompt injection.
Ingestion points: User-provided task strings are passed to the subspawn_plan.py script via the --task argument and interpolated into the generated subagent prompts.
Boundary markers: The generated prompts use a structured format with clear headers (e.g., Task:, Scope:, Mode:, Wait:) to separate the task description from operational constraints.
Capability inventory: Subagents are assigned specific roles with varying permissions, including read-only or workspace-write filesystem access, and the ability to use external research tools like GitHub and Context7.
Sanitization: While the script performs direct string interpolation, the skill mandates safety instructions across all role templates, requiring subagents to redact sensitive information (secrets, tokens, credentials) and treat the parent prompt as the authoritative source of instructions.

subspawn