report-task
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input from
$ARGUMENTSto create task files intended for consumption by a downstream agent (tasks-start). This creates a risk where malicious instructions embedded in a bug report could influence the behavior of the next agent in the pipeline. - Ingestion points: User input is ingested via the
$ARGUMENTSparameter inSKILL.mdand processed during the reproduction and writing steps. - Boundary markers: The skill does not define specific delimiters or "ignore embedded instructions" warnings when interpolating user reports into the generated markdown task files.
- Capability inventory: The skill (and the downstream agent it supports) has the capability to write files to the
.plans/tasksdirectory, executegitandgh(GitHub CLI) commands, and trigger end-to-end tests involving VM creation through the.agents/skills/e2e-test/scripts/e2eutility. - Sanitization: No sanitization, escaping, or validation of the content in
$ARGUMENTSis performed before it is used to populate the "Problem", "Changes", and "Tests" sections of the generated task files.
Audit Metadata