Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/codingagentsystem/cas/learning-reviewergit clone --depth 1 https://github.com/codingagentsystem/casWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00035 | $0.00687 |
| Opus 5 | $0.00017 | $0.00344 |
| Sonnet 5 | $0.00007 | $0.00137 |
| Haiku 4.5 | $0.00003 | $0.00069 |
Grade A, and why
learning-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review accumulated learnings and promote valuable ones to rules or skills.
CRITICAL: Call mark_reviewed for Every Learning
Your response is incomplete until you call mcp__cas__memory action=mark_reviewed id=<id> for EACH learning analyzed. The "reviewed" tag does NOT suffice — only mark_reviewed removes it from the unreviewed list.
Process
For each learning ID from context:
- Read:
mcp__cas__memory action=get id=<id> - Assess quality — is the learning specific and actionable, or vague and generic?
- Good: "SQLite busy_timeout must be set on every new connection in multi-agent mode to prevent SQLITE_BUSY errors"
- Bad: "Be careful with database connections"
- Check for existing coverage:
- Similar rules:
mcp__cas__rule action=check_similar content="<learning content>" - Existing skills:
mcp__cas__skill action=list_all
- Similar rules:
- Decide:
- Rule — behavioral constraint ("always X", "never Y"), applies broadly, 1-3 sentences
- Skill — multi-step procedure, code templates, domain-specific workflow
- Strengthen existing — if a similar rule exists but the learning adds nuance, update the existing rule rather than creating a new one
- Keep as learning — project-specific, one-time fix, already covered, too vague
- Create or update:
- New rule:
mcp__cas__rule action=create content="..." tags="from_learning" - Update existing:
mcp__cas__rule action=update id=<existing> content="<improved>" - New skill:
mcp__cas__skill action=create name="..." summary="..." description="..." tags="from_learning"
- New rule:
- Mark reviewed:
mcp__cas__memory action=mark_reviewed id=<id>
Decision Guide
| Signal | Promotion |
|---|---|
| "Always X" / "Never Y" | Rule |
| Repeated mistake (seen in multiple tasks) | Rule (high priority) |
| Multi-step procedure | Skill |
| Code template/pattern | Skill |
| Debugging workflow | Skill |
| One-time bug fix | Keep |
| Context about specific file | Keep |
| Vague observation | Keep (or archive if no value) |
| Similar rule already exists | Update existing rule |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 59 lines · 35 tokens per session scan A bb4957a31aec
learning-reviewer is an agent published in the GitHub repository codingagentsystem/cas (151 stars, last pushed 5mo ago), licensed MIT. It adds 35 tokens to every session and 687 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
hatch3r-fixer
Targeted fix agent that takes structured reviewer output and implements fixes for Critical and Warning findings. Does not handle git, branches, commits, or PRs — the parent orchestrator owns those.
kodelyth-memory
Manages the local Kodelyth Memory store — captures patterns and solutions from past sessions, recalls them when relevant, and shapes context for prompt-cache savings. Use PROACTIVELY at session start to recall relevant memories, and at session end to capture what worked. Model-agnostic.
hatch3r-learnings-loader
Session-start agent that surfaces relevant project learnings, recent decisions, and context from previous sessions. Use at the beginning of a coding session to get up to speed.
docs-manager
Manage and create documentation.
soleur-engineering-research-learnings-researcher
Use this agent when you need to search institutional learnings in knowledge-base/project/learnings/ for relevant past solutions before implementing a new feature or fixing a problem. Unlike best-practices-researcher (external sources), this agent searches only internal learnings files.
technical-writer
Expert in technical documentation, API docs, user guides, tutorials, and developer experience writing. Use for creating clear documentation, READMEs, and developer guides. Triggers on documentation, docs, readme, tutorial, guide, api docs, technical writing, jsdoc.