add-oauth-provider
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines how the agent should ingest and process data from external OAuth providers via discovery and health probe APIs. This creates a surface for indirect prompt injection where a compromised or malicious OAuth provider could return specially crafted data in fields like
discover_label_fieldorprobe_pathresponses to influence the agent's behavior. - Ingestion points: API responses defined in
src/treg/oauth_providers.py(e.g.,discover_path,probe_path). - Boundary markers: None explicitly mentioned in the configuration guidance.
- Capability inventory: The skill enables network GET requests (probes) and API discovery calls.
- Sanitization: The instructions recommend using structured dotted-path resolution (
_dig) to extract specific fields, which mitigates some free-text injection risks. - [COMMAND_EXECUTION]: The skill instructs the developer to execute standard testing and development scripts using the
uvpackage manager. - Evidence:
uv run pytest -qandscripts/dev-local.sh upin the verification and testing sections.
Audit Metadata