apod-to-short
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data by fetching NASA APOD explanations and titles using the
WebFetchtool. An attacker who controls the content of a target URL could attempt to embed instructions to influence the agent's output. - Ingestion points: External content is fetched in Workflow Step 1 (SKILL.md) and the
apodinput (manifest.json). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined for the processed text.
- Capability inventory: The skill uses
WebFetchandReadtools. It lacks high-risk capabilities such as file system writing, shell command execution, or network POST operations. - Sanitization: No explicit sanitization logic is present, though the workflow's focus on extracting a single fact and adhering to specific word counts acts as a natural constraint.
- [EXTERNAL_DOWNLOADS]: The skill is configured to use the
WebFetchtool to retrieve data from external APOD sources. This is the intended primary function of the skill and is documented neutrally.
Audit Metadata