The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the merging of multiple external prose files into a single document, which creates an indirect prompt injection surface. Malicious instructions embedded in any individual section could influence the behavior of downstream AI agents (such as the 'voice-gate' mentioned in the documentation) or mislead human reviewers.
Ingestion points: Reads content from sections/*.md, outline/transitions.md, outline/tables_appendix.md, and GOAL.md.
Boundary markers: The skill uses standard Markdown headings to separate merged sections but does not implement explicit boundary markers or 'ignore' instructions to prevent the agent from following commands embedded in the merged prose.
Capability inventory: The script is limited to file system read/write operations within the provided workspace via the pathlib and tooling.common libraries. It does not possess network access or shell execution capabilities.
Sanitization: While the script performs structural cleanup on table syntax and heading normalization, it does not sanitize or filter the actual prose content for instructions, control characters, or potential injection patterns.

section-merger