terra-planned-workouts

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to install vendor-specific tools via npm install -g @tryterra/cli or brew install tryterra/tap/terra to facilitate integration management.
  • [COMMAND_EXECUTION]: The agent is encouraged to use a CLI tool (terra) to perform administrative tasks autonomously rather than requesting user intervention.
  • [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to fetch and process live documentation from docs.tryterra.co at runtime, which serves as a potential vector for instructions hidden in external data. 1. Ingestion points: SKILL.md (instructions to fetch garmin-exercise-reference.md and sport-specific-examples.md). 2. Boundary markers: Absent; no instructions are provided to the agent to treat external doc content as data only. 3. Capability inventory: The agent can use curl and the terra CLI to interact with external services and the local system. 4. Sanitization: Absent; the skill does not specify validation or sanitization of the fetched markdown files.
  • [PROMPT_INJECTION]: The instructions include a directive to "Reach for it [the CLI] instead of handing the task back to the user," which explicitly overrides the default agent behavior of deferring to the user for administrative actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 09:04 AM
Security Audit — agent-trust-hub — terra-planned-workouts