Agent Memory Reflector

Enables AI agents to review past decisions, identify reasoning loops, and produce insights for self-improvement to enhance their cognitive processes.

Security Scan

The skill's code and instructions are consistent with a local-memory reflection tool that stores and analyzes agent prompts/responses on disk; it does not request credentials or reach out to external endpoints, but it will persist potentially sensitive conversation data locally.

Purpose & Capability

The name and description (reflect on past decisions, detect loops, generate suggestions) match the provided implementation: a small local logger/analyzer that writes to .agent_memory and produces reflection reports. Minor inconsistencies in bookkeeping names: SKILL.md examples call python agent_memory_reflector.py while the included script is tool.py (module docstring/header also uses a different filename/version), but this looks like sloppy naming rather than malicious mismapping.

Instruction Scope

Runtime instructions and the CLI usage map to the code. The tool logs entire prompts/responses and metadata to a local .agent_memory/memory.jsonl and writes reflection reports to .agent_memory/reflections.jsonl. The instructions do not ask the agent to read unrelated system files or environment variables. Note: storing full prompts/responses may persist sensitive data and the SKILL.md does not warn about that.

Install Mechanism

There is no external install script or downloads—no network fetches or archive extraction. The skill is delivered as source (tool.py) and will run locally. No package managers or remote URLs are used.

Credentials

The skill declares no required environment variables, credentials, or config paths and the code does not access env vars or external services. This is proportionate to its stated purpose. The only resource it uses is the local filesystem (a directory named .agent_memory).

Persistence & Privilege

The skill does persistent local storage in .agent_memory and appends logs/reports there. It does not set always:true and does not modify system/other-skill configs. Be aware that if an agent invokes this skill autonomously, it could cause repeated writes of conversation history without additional prompts.

Assessment

This skill appears to be what it claims: a small local reflection logger/analyzer. Before installing: - Expect it to store full prompts, responses, and supplied metadata in plaintext under .agent_memory in the working directory; do not run it where prompts/responses contain secrets you cannot store. - The SKILL.md references agent_memory_reflector.py but the provided file is tool.py—verify filenames before running and inspect the source (tool.py) yourself. - There are no network calls or credential requests in the code, which reduces exfiltration risk; nevertheless, review the code if you will run it on sensitive data. - If you need privacy: modify the code to encrypt logs, change the storage path, or sanitize/redact sensitive fields before logging; consider running the skill inside an isolated container or environment. - The listed price ($29) in SKILL.md is unrelated to technical behavior; verify licensing/purchase outside this package if relevant.

A minimal, embeddable reflection engine that gives AI agents the ability to examine their own past decisions, detect reasoning loops, and generate actionable self-improvement insights—like a debugger for agent cognition.

Usage

Log an agent interaction:

python agent_memory_reflector.py --agent "task_planner_v3" \
  --prompt "How should I deploy the microservice?" \
  --response "You can use Kubernetes with Helm." \
  --meta '{"confidence":0.8, "retrieved_context":true}'

Generate a reflection report:

python agent_memory_reflector.py --agent "task_planner_v3" --reflect

Price

$29.00