openspec-trace-logger
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local CLI tools including
codegraph,rg(ripgrep),adb(Android Debug Bridge), and./gradlew(Gradle Wrapper) to discover call chains, search source code, capture logs, and verify builds. These operations are performed within the local development environment. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from bug descriptions and existing project source code, which serves as a surface for indirect prompt injection.
- Ingestion points: User input through the
AskUserQuestiontool and reading of local*.ktand*.javasource files. - Boundary markers: The skill does not define explicit delimiters to separate user-provided behavior descriptions from the agent's internal diagnostic logic.
- Capability inventory: Extensive file-system write access (for source code instrumentation), shell command execution (
adb,gradle), and project-wide code search (rg). - Sanitization: The skill contains strict internal instructions to never log PII, mask sensitive values like tokens or passwords, and use hash codes for identification, which reduces the potential impact of data exposure.
- [DYNAMIC_EXECUTION]: The skill modifies existing application source code at runtime to insert logging statements. Although these modifications follow predefined templates, the act of programmatically altering executable source files represents a dynamic execution vector. This risk is managed by a mandatory cleanup phase that requires build verification and the removal of all instrumentation markers before the session ends.
Audit Metadata