The Agent Skills Directory

[SAFE]: The skill coordinates the execution of established development skills (manifest-dev:figure-out, manifest-dev:define, and manifest-dev:do) to automate planning and implementation tasks.
[SAFE]: The skill includes a domain validation check for the --babysit argument, ensuring that external pull request data is only fetched from github.com, a well-known and trusted service.
[PROMPT_INJECTION]: The skill utilizes a "full-chain goal-setting backstop" to provide the agent with a detailed framework for autonomous operation. These instructions are designed to enforce rigorous evidence gathering and quality control throughout the task lifecycle and do not attempt to bypass core safety guardrails.
[PROMPT_INJECTION]: The skill ingests user-defined tasks and external Pull Request content, which acts as a potential surface for indirect prompt injection. However, this is inherent to the skill's purpose as an orchestration layer.
Ingestion points: Task descriptions via $ARGUMENTS and GitHub PR content via the --babysit flag.
Boundary markers: The skill does not explicitly utilize delimiters or instructions to ignore embedded commands within the ingested text.
Capability inventory: The skill chains to manifest-dev:do, which typically possesses filesystem and command execution capabilities to perform development work.
Sanitization: Ingested content is passed to sub-skills without additional filtering or sanitization.

auto