The Agent Skills Directory

[PERSISTENCE]: The scripts/install-cli.mjs script installs a persistent wrapper for the agent-journal CLI in the user's ~/.local/bin directory. This is a standard setup procedure for local tools to ensure the command remains available across terminal sessions and for automated hook execution.
[INDIRECT_PROMPT_INJECTION]: The skill implements 'Authoring Floors' which generate dynamic prompts for the agent based on session activity and tool observations.
Ingestion Points: Data is ingested from the observation plane, specifically tool_call and tool_result events processed by the floor command.
Boundary Markers: The generated prompt text does not use specific delimiters to separate observed tool data from instructions.
Capability Inventory: The skill is granted Read, Write, Grep, Glob, and Bash permissions to manage local journal files and execute the provided CLI scripts.
Sanitization: The CLI includes a dedicated redaction engine (src/redact.ts) that uses regular expressions to mask common secret patterns (API keys, tokens, JWTs, and private keys) before they are committed to the journal storage.
[EXTERNAL_DOWNLOADS]: The documentation in references/adapters.md refers to the author's GitHub repository at github.com/crissmoldovan/agent-skills to obtain adapters for various agent harnesses. This is documented as a vendor-provided resource for expanding the skill's capabilities.
[COMMAND_EXECUTION]: The CLI tool performing the journaling logic uses standard file system operations to manage local .jsonl segment files within the ~/.agents/journal/ directory. The installation script also utilizes chmod and rename to set up the CLI wrapper locally.

decision-journal