The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform remote operations using ssh and rsync to interact with a target Docker host (e.g., ssh docker-host 'cd ~/docker-compose/project && docker compose build <service>' in references/remote-docker-verification.md). it also uses docker compose and git commands extensively.- [DYNAMIC_EXECUTION]: The pipeline requires the creation and execution of a 'reproduction script' (shell script or Dockerfile) in Phase 1 to verify bug reports before proceeding with fixes. This involves generating and running code based on external issue descriptions.- [REMOTE_CODE_EXECUTION]: The combination of generating reproduction scripts and the ability to execute commands on a remote Docker host via SSH constitutes a remote code execution capability.- [INDIRECT_PROMPT_INJECTION]:\n
Ingestion points: The skill ingests untrusted data from issue URLs, issue descriptions, and repository-level files such as CONTRIBUTING.md, AGENTS.md, and AI_POLICY.md (SKILL.md, trigger-patterns.md).\n
Boundary markers: Absent; the instructions do not specify the use of delimiters or 'ignore' instructions for the external content processed during the orientiation and comprehension phases.\n
Capability inventory: The agent has extensive capabilities including file system access (read_file, write_file, patch), network operations (git, rsync, ssh), and command execution (docker compose).\n
Sanitization: While the skill includes a security gate (Phase 3.5) to scan the final contribution for secrets and vulnerabilities, it lacks input sanitization for the data processed during the initial comprehension phases.- [CREDENTIALS_UNSAFE]: The file references/gate-criteria.md contains a hardcoded dummy AWS credential string (AWS_KEY=AKIA...) used as part of a falsification test to verify that the agent's security scanner is functional.

contribution-pipeline