The Agent Skills Directory

[COMMAND_EXECUTION]: The provided utility script render-graphs.js uses child_process.execSync to invoke the system dot command for rendering Graphviz diagrams. This is used for visualizing protocol flows.
[DYNAMIC_CONTEXT_INJECTION]: SKILL.md contains examples and instructions for using the !command syntax (e.g., !git log, !gh pr diff). This feature executes shell commands when the skill is loaded by the agent environment. The documented examples are standard developer tools.
[DYNAMIC_EXECUTION]: The render-graphs.js script dynamically extracts and executes Graphviz code blocks from markdown files at runtime using the dot utility.
[INDIRECT_PROMPT_INJECTION]: The skill's core methodology involves running 'pressure scenarios' with subagents and incorporating their output (failures and rationalizations) into the protocol documentation. This creates a surface where malicious instructions could potentially be ingested from untrusted subagent responses.
Ingestion points: Failures and rationalizations recorded from subagent interactions are intended to be authored into SKILL.md protocol files.
Boundary markers: No specific delimiters or safety warnings for handling untrusted subagent output are prescribed.
Capability inventory: The skill possesses command execution capabilities through render-graphs.js and dynamic loading features via the platform's !command syntax.
Sanitization: The skill lacks explicit instructions for sanitizing or escaping content derived from external subagent sessions before interpolation into documentation.

protocol-authoring