The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes 'PRD shards' (Markdown files) to generate implementation epics. If these shards contain malicious instructions or hidden payloads, they could influence the agent's behavior during document generation.
Ingestion points: Shard files located in the PRD root directory (e.g., docs/prd/{domain}/prd.{P}.{feature}/).
Boundary markers: The skill relies on standard Markdown structure and specific frontmatter fields but lacks explicit delimiters or instructions to ignore embedded commands within the shards themselves.
Capability inventory: File system access (read/write) via local Node.js scripts and shell script execution.
Sanitization: The skill performs basic regex-based cleaning on titles and unquotes YAML values in the generate-prd-epic-index.mjs script.
[COMMAND_EXECUTION]: The skill executes locally bundled Node.js and Bash scripts (doc-id.js, generate-prd-epic-index.mjs, read-config.sh) to perform document ID allocation, roadmap generation, and configuration parsing. These operations are restricted to local file manipulation and do not involve network connectivity.
[DYNAMIC_EXECUTION]: The configuration utility read-config.sh executes a hardcoded Python script via python -c to parse YAML files. The Python script includes security hardening to prevent local module shadowing by explicitly filtering sys.path.

create-epics-from-shards