find-funding-opportunities
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill performs network requests using
curltogapapi.karmahq.xyz. This is a vendor-owned domain used to fetch grant and hackathon program data, which is the primary intended function of the skill. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party data (titles, descriptions, and metadata) returned from the Karma API. It mitigates the risk of instructions embedded in this data by:
- Establishing clear ingestion points from the API responses in
SKILL.md. - Using explicit boundary markers like
[Begin program data]and[End program data]to isolate external content. - Providing a dedicated 'Content Safety' section that instructs the agent to treat all fetched fields as display-only and to ignore any text resembling directives or prompt injections.
- Limiting capabilities to basic read and format operations, with no file-write or subprocess execution on the fetched content.
- [COMMAND_EXECUTION]: The skill uses the shell commands
uuidgenandcurlto manage unique session identifiers and API interactions. These are standard utility calls necessary for the functionality described in the workflow.
Audit Metadata