The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core function involves reading and updating project instructions, which could allow malicious content from untrusted files to influence agent behavior.
Ingestion points: Consumes documentation from AGENTS.md, .codex/rules/, and native agent/skill configurations (TOML and Markdown) as identified in SKILL.md and references/sync-contract.md.
Boundary markers: Strictly limited to an explicitly authorized inventory of paths and respects user-defined scope narrowing to prevent unauthorized file manipulation.
Capability inventory: Performs filesystem discovery using the rg command and executes file modifications via the apply_patch tool within the git-tracked environment.
Sanitization: Employs independent read-only post-sync checks and audits native data structures (TOML/YAML) to ensure integrity and compliance with active policies.

memory-sync