The Agent Skills Directory

[Knowledge Ingestion Surface]: The skill involves the agent reading and interpreting content from markdown files within the repository's learnings/ directory to guide its behavior.
Ingestion points: Content is retrieved from {{BASE_PATH}}/learnings/<role>/<slug>.md as described in SKILL.md.
Boundary markers: While the skill does not use explicit character-based delimiters, it implements a procedural boundary by instructing the agent to scan the first three lines for relevance before reading the full file body.
Capability inventory: The skill utilizes standard file system operations (list_dir, read_file) and does not require network access or privileged command execution.
Sanitization: The skill relies on the structured markdown format to organize information, though it does not specify programmatic sanitization of the narrative text.
[Persistence Pattern]: The skill documentation describes the learnings as "persistent across runs." This refers to standard project documentation stored in Git-tracked files intended for human and agent collaboration, rather than a hidden or unauthorized persistence mechanism.

sharing-learnings