The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Python scripts for document processing and environment configuration.
SKILL.md directs the agent to execute scripts/sync_tech_specs.py and scripts/install_project_hook.py using the system Python interpreter.
The scripts use the argparse module for safe argument handling and pathlib for secure path management.
[PERSISTENCE]: The skill configures project-specific hooks to ensure documentation remains synchronized across agent sessions.
scripts/install_project_hook.py modifies the .codex/hooks.json configuration file to register PostToolUse and Stop events.
These hooks are designed to automatically trigger the synchronization script whenever relevant files are modified, ensuring the HTML representation stays current.
The skill explicitly instructs the agent to ask the user to review and trust these hooks through the platform interface.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown content to generate HTML files, with mitigations implemented to prevent injection.
Ingestion points: Markdown source files are read from the docs/tech-specs/ directory within the repository.
Boundary markers: The process uses structured YAML frontmatter (delimited by ---) and specific Markdown headings to segment content.
Capability inventory: The skill has capabilities for file system writes and automated command execution via project-local hooks.
Sanitization: The rendering engine in scripts/sync_tech_specs.py performs thorough HTML escaping on all inline content and code blocks, and utilizes a safe_link validator to prevent the inclusion of dangerous URI schemes such as javascript:.

global-tech-spec