writing-skills
Warn
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The files 'persuasion-principles.md' and 'testing-skills-with-subagents.md' explicitly instruct the user to use 'Authority', 'Scarcity', and 'Commitment' principles to ensure agent compliance. It advocates for the use of imperative markers such as 'YOU MUST', 'Never', 'Always', and 'No exceptions'. These patterns are designed to override an agent's internal reasoning or decision-making process, which is characteristic of bypass-oriented prompt instructions.
- [COMMAND_EXECUTION]: The skill includes an executable Node.js script 'render-graphs.js' that utilizes 'child_process.execSync' to run system-level commands, specifically 'dot' (Graphviz) and 'which dot'. This allows the skill to spawn subprocesses on the host machine to generate SVG files from markdown blocks.
- [DATA_EXPOSURE]: The 'render-graphs.js' script takes a directory path as a command-line argument and performs file read and write operations within that path. While localized to rendering graphs, it represents a capability to interact with the local filesystem and write data to computed paths.
Audit Metadata