The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources which can be influenced by external inputs or previous malicious interactions.
Ingestion points: The skill reads data from .learnings/LEARNINGS.md, ERRORS.md, FEATURE_REQUESTS.md, HEALS.md, and session traces via the entire CLI tool, as detailed in Step 1 and the Session Trace Analysis section of SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or delimiters for the summaries and details extracted from logs when producing the Gap Report.
Capability inventory: The resulting Gap Report is used by the harness-updater agent to modify project instruction files (such as CLAUDE.md and AGENTS.md) and by the eval-creator skill to generate test cases.
Sanitization: There is no documented sanitization or validation process for the 'Summary' or 'Details' fields extracted from log entries before they are included in the promotion recommendations.
[COMMAND_EXECUTION]: The skill utilizes the entire command-line interface to perform deep trace analysis of session history.
Evidence: SKILL.md specifies the use of shell commands to query session data, including entire --version, entire rewind --list, and entire explain with various flags for checkpoint and transcript retrieval.

learning-aggregator