The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interprets pull request content to decide if visual changes match developer intent, which presents a surface for indirect prompt injection from untrusted PR data.
Ingestion points: Pull request titles, descriptions, linked issues, and code diffs are ingested by the agent in SKILL.md (Workflow step 3).
Boundary markers: No specific delimiters or instructions to ignore embedded commands are mandated for the processing of external pull request content.
Capability inventory: The skill has the capability to execute shell commands via the argos CLI, including creating comments and submitting reviews on the repository.
Sanitization: There are no specified sanitization or validation requirements for handling the PR data before it influences the agent's review logic.
[COMMAND_EXECUTION]: The skill relies on the execution of the argos command-line tool to perform its primary visual testing review functions, which involves handling authentication tokens as command parameters.
Evidence: Multiple shell command patterns are provided in SKILL.md, including argos build get, argos build snapshots, argos comment create, and argos review create. The skill includes a security rule: "never print token values."

argos-pr-review