skills/j5ik2o/takt-sdd/takt-optimize/Gen Agent Trust Hub

takt-optimize

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes a shell script scripts/validate-takt-files.sh to verify optimized files. This script contains a relative path reference ../../takt-piece/scripts/validate-takt-files.sh, which relies on a specific external directory structure being present in the agent's environment.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to its data ingestion and file generation capabilities.
  • Ingestion points: In Step 1 and Step 7, the skill reads untrusted data from .takt/logs/*.jsonl (execution logs) and user-provided piece/facet files located in ~/.takt/pieces/ or .takt/pieces/.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded directives within the ingested logs or facets are specified in the prompt logic.
  • Capability inventory: The skill possesses the ability to create, delete, and modify files (Step 3) and execute shell scripts (Validation step).
  • Sanitization: The skill lacks a sanitization layer for the ingested data; instructions from the logs or existing facets could potentially influence the logic of the newly generated files or the parameters of the validation command.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 04:12 AM