The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation in references/skill-routing.md and references/project-md-template.md recommends executing an unpinned third-party package from the npm registry via npx @stevestomp/ohno-cli. Dynamic execution of unpinned packages can pose supply chain risks if the third-party package is compromised.
[COMMAND_EXECUTION]: The helper scripts provided in references/skill-routing.md execute shell commands using npx combined with piped inline Node.js evaluation commands (node -e). This pattern relies heavily on the runtime integrity of the external tool.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to parse and process untrusted external text inputs, specifically Product Requirement Documents (PRDs) and concept briefs, exposing it to indirect prompt injection vulnerabilities.
Ingestion points: Processes external files through the [path-to-prd] argument specified in SKILL.md.
Boundary markers: Absent; there are no explicit delimiters or strict boundary markers defined to isolate user text from the model instructions.
Capability inventory: The skill possesses the capability to modify and create project hierarchies (epics, stories, tasks) in the local tracking system via MCP tools.
Sanitization: No sanitization or safety filters are applied to the ingested PRD data before processing.

planning