The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a hierarchical configuration loading pattern that ingests data from AGENTS.md files found in the workspace and parent directories. This introduces a risk of indirect prompt injection, where an adversary could place a malicious instruction file in a directory targeted by the scan.
Ingestion points: AGENTS.md or agents.md files located via directory traversal and Glob operations (SKILL.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' directives when merging external content into the unified context.
Capability inventory: The skill uses Read, Write, Edit, Grep, Glob, and Bash tools (SKILL.md).
Sanitization: Absent. The instructions focus on merging and reporting conflicts rather than validating or escaping the content of the discovered files.
[EXTERNAL_DOWNLOADS]: The skill contains references to documentation hosted on Anthropic's official domain (docs.anthropic.com) and the author's infrastructure (jeremylongshore.com, intentsolutions.io). These references are for documentation and setup purposes and are associated with trusted or vendor-related sources.
[COMMAND_EXECUTION]: Includes a shell script (scripts/check-agents-md.sh) designed to verify the presence of an AGENTS.md file and notify the user. The script performs local file checks and prints status messages without executing remote code or performing dangerous filesystem operations.

agent-context-loader