The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill reads spreadsheet values and Drive comments via the google_workspace_request and drive.comments.list methods in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the ingested spreadsheet content.
Capability inventory: The skill can execute various deepline tool commands, perform file system writes (mkdir, tee), and fetch run data, which could be leveraged if the agent is manipulated by injected data.
Sanitization: The skill lacks explicit sanitization or validation steps for the content retrieved from Google Sheets before it is used to revise Plays or set expectations.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the deepline CLI tool from a custom NPM registry.
Evidence: npm install -g deepline --registry https://code.deepline.com/api/v2/npm/ in SKILL.md.
Context: The download targets a vendor-specific registry associated with the primary functionality of the skill.
[COMMAND_EXECUTION]: The skill relies on several shell commands to manage the environment and data flow.
Evidence: Execution of npm, deepline, jq, node, mkdir, and tee is defined throughout SKILL.md to facilitate the review loop, data transformation, and directory management.

deepline-plays-review