The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/brainerd-claude.sh, scripts/brainerd-codex.sh) and Windows batch files (scripts/brainerd-claude.cmd, scripts/brainerd-codex.cmd) to manage memory persistence routing.
[EXTERNAL_DOWNLOADS]: The provided scripts use npx --yes tsx to run local TypeScript source files. This command fetches the tsx package from the official NPM registry if it is not already present in the local cache.
[COMMAND_EXECUTION]: The skill generates temporary JSON payloads and writes them to /tmp/brainerd-reflect.json as an intermediate step before applying updates to the repository.
[PROMPT_INJECTION]: The skill processes untrusted session data to identify durable learnings, which introduces a surface for indirect prompt injection.
Ingestion points: Conversation history is ingested via brainerd_current_session, and existing repository memory is read from brain/index.md and brain/principles.md.
Boundary markers: The instructions specify distilling content into small changes and require a specific Brainerd summary: prefix for the final output.
Capability inventory: The skill possesses the capability to modify repository files through local CLI wrappers.
Sanitization: No explicit sanitization or escaping of session content is performed; the skill relies on the agent's summarization logic to filter inputs.

brainerd-reflect