relay-ready
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user request text to judge readiness and generate handoff artifacts for downstream automated processes.
- Ingestion points: Untrusted data enters the agent context via the
request_textandhandofffields in the contract file located at/tmp/relay-ready-contract.json. - Boundary markers: Instructions in
SKILL.mdadvise the agent to judge readiness using only text outside fenced code blocks, serving as a basic structural delimiter. - Capability inventory: The skill has the capability to write immutable request bundles to the user's home directory (
~/.relay/requests/), which are subsequently used as the source of truth for planning and execution phases. - Sanitization: While the scripts perform strict schema validation and identifier safety checks for file system operations, they do not sanitize or filter the natural language content of the intake requests.
- [COMMAND_EXECUTION]: The skill executes the
gitcommand locally viaexecFileSyncto determine repository paths and generate unique storage identifiers. This operation is scoped to the target repository provided via CLI arguments.
Audit Metadata