The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and follow instructions from TASKS.md files and  HTML comments found within the repository. These external data sources can be used to influence the agent's autonomous behavior.
Ingestion points: The agent reads TASKS.md (specifically fields like **Details**, **Acceptance**, and **Research**) and repository-wide policy comments in SKILL.md and references/enrich-blocked-tasks.md.
Boundary markers: While the skill uses structured markdown metadata, it does not explicitly instruct the model to ignore or sanitize natural language instructions embedded within these fields.
Capability inventory: The skill has access to powerful tools including Bash, Write, Edit, and MultiEdit, which allows for arbitrary file modifications and system commands.
Sanitization: No sanitization, escaping, or validation of the content within TASKS.md or policies is performed before the agent acts on the information.
[COMMAND_EXECUTION]: The skill includes a pre-flight check that automatically executes a local script at scripts/check-zero-ship-streak.mjs using node. This behavior relies on the presence of the file in the repository and executes it to determine whether the autonomous session should proceed. Additionally, the skill extensively utilizes the git and gh (GitHub) CLI tools to manage branches, commits, and pull requests.
[EXTERNAL_DOWNLOADS]: The skill performs network operations using git pull --rebase and gh pr list to synchronize the local environment with remote GitHub repositories. These operations target well-known services (GitHub) and are part of the skill's core functionality for repository maintenance.

next-task