skillify
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web content (HTML, page text, snapshot trees) to synthesize scraping scripts, creating a vulnerability surface for indirect prompt injection where malicious instructions in a web page could influence the generated code.
- Ingestion points: Scraped content from
aside replandaside exectools. - Boundary markers: The skill includes explicit instructions to ignore instructions found in page content and report them as potential injection attempts.
- Capability inventory: Shell access (
Bash), file system modification (Write,commitSkill), and code execution (bun test). - Sanitization: Instructions mandate treating all extracted strings as attacker-influenceable during the code synthesis process.
- [DYNAMIC_EXECUTION]: The skill generates TypeScript source files (
script.ts) and test suites (script.test.ts) at runtime, which are subsequently executed using thebunruntime for verification. It also dynamically resolves SDK paths from the local environment to bundle them with the new skill. - [COMMAND_EXECUTION]: The skill uses several local binaries (
gstack-skill-start,gstack-skill-end,gstack-question-log,gstack-learnings-log) to manage session state, telemetry, and preference logging. These are documented as part of the vendor's framework infrastructure.
Audit Metadata