The Agent Skills Directory

[SAFE]: The skill performs automated project bootstrapping by creating AGENTS.md and CLAUDE.md files. This behavior is transparent, documented, and restricted to the workspace root.
[SAFE]: The skill uses official Adobe documentation and repository links (e.g., experienceleague.adobe.com, github.com/adobe/*), which are trusted vendor resources.
[SAFE]: File operations are restricted to reading project configuration (pom.xml, package.json) and writing new documentation files. It explicitly refuses to overwrite existing files, preventing accidental data loss.
[INDIRECT_PROMPT_INJECTION]: The skill has a minor surface for indirect prompt injection as it ingests untrusted metadata (project names and module lists) from files like pom.xml and interpolates them into agent-facing instructions (AGENTS.md).
Ingestion points: pom.xml (<name>, <artifactId>, <modules>) and ui.frontend/package.json.
Boundary markers: Absent; the content is placed directly into the generated markdown.
Capability inventory: Uses fs.write to create permanent project documentation that influences future agent behavior.
Sanitization: Includes basic string manipulation (title-casing, prefix stripping) which provides minimal protection against malicious payloads in project metadata.
[COMMAND_EXECUTION]: While the generated AGENTS.md file contains examples of Maven and npm build commands, the skill itself does not execute these commands; it only documents them for the developer.

ensure-agents-md