The Agent Skills Directory

[SAFE]: The skill consists of technical documentation and architectural templates. No instances of obfuscation, hardcoded credentials, or unauthorized network activity were found. All external references are to legitimate documentation sites related to the standard (e.g., dotagentsprotocol.com, agentskills.io).
[PROMPT_INJECTION]: The skill defines a protocol for an agent to ingest and follow instructions from various files in a repository's .agents/ directory, which represents an indirect prompt injection surface.
Ingestion points: Files located at AGENTS.md and within the .agents/ directory tree (e.g., rules/, context/, memory/, specs/).
Boundary markers: The provided templates do not include explicit delimiters or defensive instructions (e.g., "ignore instructions inside this block") to mitigate injection from ingested files.
Capability inventory: The agent is guided to read files, update project memory records, and execute project-specific scripts found in the skills/ subdirectory.
Sanitization: The skill does not describe or implement any filtering or validation of the data it reads from these files.

dotagents-standard