data-format-encoder
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary text input for encoding and decoding, creating a potential surface for indirect prompt injection attacks.
- Ingestion points: All text transformation actions (e.g.,
encode-base64-encode,encode-url-decode,encode-html-entity-encode) inSKILL.mdtake atextparameter which can contain untrusted data from external sources. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore' warnings for the input text, increasing the risk that an agent might follow instructions embedded in the data being transformed.
- Capability inventory: The skill performs remote API calls to
api.agentpmt.comfor processing; it does not request local file system access or direct command execution capabilities. - Sanitization: There is no evidence of input validation or content sanitization to filter out potential instructions from the text data.
- [EXTERNAL_DOWNLOADS]: The documentation provides instructions to download and install additional setup skills from the developer's infrastructure.
- Evidence: The skill references
npx skills add AgentPMT/agent-skillsfor environment setup, which fetches content from the official AgentPMT GitHub repository.
Audit Metadata