The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, which constitutes an attack surface for indirect prompt injection.
Ingestion points: The skill reads data from GitHub Pull Requests, including comments, review threads, and bot summaries (defined in references/repository-playbook.md).
Boundary markers: The skill explicitly establishes 'security boundaries' and 'landing boundaries' in SKILL.md and references/security-boundaries.md. It uses a 'context gate' logic to classify automated feedback.
Capability inventory: The skill has the capability to route tasks to executors with write authority on GitHub and local runtimes, as defined in references/capability-registry.yaml and agents/roles.yaml.
Sanitization: The skill mandates classification of external inputs (e.g., CURRENT_ACTIONABLE, ALREADY_FIXED) and verification against current code/tests before acting on external feedback, which serves as a reasoning-based sanitization layer.
[SAFE]: The skill follows security best practices by implementing 'fail-closed' rules and prohibiting the storage of secrets or credentials within its metadata or reference files, as stated in references/security-boundaries.md.
[SAFE]: External references to agentskills.io and agent-plugins.org are used for technical specifications and do not involve executable code or untrusted package installations.

gtd-capability-router