The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes bash through the agent's toolset to perform system audits and file management tasks during the analysis and implementation phases.
It employs commands such as find, grep, ls, and wc to catalog notes, MOCs (Maps of Content), templates, and existing hooks.
It uses git mv to rename and restructure directories while attempting to preserve file history.
It executes a local shell script (validate-kernel.sh) located within the plugin root to verify the integrity of the re-derived system.
[DYNAMIC_EXECUTION]: The skill is designed to modify and generate executable scripts and instruction files as part of its restructuring process.
In Phase 6, it modifies skill instruction files (SKILL.md) within the .claude/skills/ directory to update vocabulary.
It adds or removes shell scripts (.sh) within .claude/hooks/ based on changes to the system's automation level.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from external sources and user notes to drive its re-derivation logic, creating a potential vector for indirect influence.
Ingestion points: The skill reads ops/derivation.md, ops/health/ reports, ops/observations/ friction patterns, and user-authored content in the notes/ directory.
Boundary markers: The instructions include a strict "ABSOLUTE INVARIANT" regarding content preservation and require user approval before implementation, but there are no specific markers to delineate or sanitize instructions potentially embedded within the notes being analyzed.
Capability inventory: The skill has access to Bash for command execution and Write/Edit tools for modifying system configurations, templates, and hooks.
Sanitization: No explicit sanitization or validation of the content extracted from notes or observation logs is performed before that data is used to inform architectural changes.

reseed