The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content from PR discussions and review findings, which creates a surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent context through PR comments, findings from the sanctum:pr-review dependency, and participant metadata as described in SKILL.md and modules/capture-workflow.md.
Boundary markers: The skill lacks explicit boundary markers or instructions to ignore embedded commands within the ingested PR content before storage.
Capability inventory: The skill possesses the capability to execute a local management script (scripts/palace_manager.py) and perform file-write operations to update the memory palace repository.
Sanitization: There is no evidence of sanitization, escaping, or validation of the PR discussion text within the provided modules.
[COMMAND_EXECUTION]: The skill uses a local script to manage and search the captured knowledge base.
Evidence: The skill references executing python scripts/palace_manager.py for searching and exporting entries in SKILL.md and modules/search-patterns.md.

review-chamber