phaser-feedback
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted feedback from external sources like Reddit and Discord to generate code.\n
- Ingestion points: Step 1 in
SKILL.mdandexamples/worked-session.mdexplicitly instructs the agent to take feedback verbatim from these platforms.\n - Boundary markers: There are no specific instructions or delimiters provided to isolate the untrusted feedback text from the agent's logic during the code generation phase.\n
- Capability inventory: The skill generates and executes JavaScript scenarios using a Node.js harness (
playtest.mjs) that supportsevalblocks.\n - Sanitization: No sanitization logic is provided to filter malicious instructions within player feedback, relying entirely on the agent's interpretation.\n- [DYNAMIC_EXECUTION]: The skill's workflow involves the automated creation and execution of scripts.\n
- Script Generation: The agent creates
.mjsfiles withevalblocks based on interpreted feedback, as seen inSKILL.md(Step 4) andreferences/feedback-to-scenario.md.\n - Execution: These files are run locally via
node, providing a path for agent-generated (and potentially attacker-influenced) code to execute in the user's environment.
Audit Metadata