The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a self-evolution pattern, instructing the agent to programmatically update the SKILL.md file's taxonomy and the references/evolution-log.md file when new failure modes or resurrection triggers are encountered. This functional self-modification is limited to documented project metadata.
[INDIRECT_PROMPT_INJECTION]: The skill includes a process for scanning and reviewing archived research files to trigger the resurrection of dormant ideas. This represents a potential vulnerability surface where malicious instructions embedded in research data could influence the agent's decision-making process.
Ingestion points: Reads files from the references/archive/ directory and exhumation-log.jsonl.
Boundary markers: Relies on YAML frontmatter and Markdown structure within archived files to separate metadata from content.
Capability inventory: Uses Read, Write, Edit, Grep, and Glob tools to manage the archive and update logs.
Sanitization: There is no evidence of specific sanitization or filtering logic to prevent the execution of instructions found within archived data.

crucible-emergent-resurrection