Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add ShreyPaharia/octomux/plugin install octomuxWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/shreypaharia/octomux/review-learnings)<a href="https://agentmods.dev/skills/shreypaharia/octomux/review-learnings"><img src="https://agentmods.dev/badge/skills/shreypaharia/octomux/review-learnings/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/shreypaharia/octomux/review-learnings"><img src="https://agentmods.dev/badge/skills/shreypaharia/octomux/review-learnings.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00041 | $0.02275 |
| Opus 5 | $0.00020 | $0.01137 |
| Sonnet 5 | $0.00008 | $0.00455 |
| Haiku 4.5 | $0.00004 | $0.00228 |
Grade B, and why
review-learnings scanned grade B with 2 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 12d ago.
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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
cat ~/.claude/projects/<project-key>/<session-id>.jsonl Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s http://localhost:7777/api/tasks | jq . How it starts
The opening of the file, as written. The whole thing — 264 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review Learnings
Extract reusable learnings from closed octomux-agents tasks and add them to the repo's
.claude/rules/learnings.md so future agents (and human sessions) benefit automatically.
Usage
/review-learnings— review all unreviewed closed tasks/review-learnings <task-id>— review a specific task/review-learnings add "..."— manually add a learning to current repo
Manual Add Mode
If the argument starts with add, extract the quoted text and append it as a bullet to
.claude/rules/learnings.md in the target repo:
-
Determine target repo:
- If
--repo <path>is provided, use that path - Otherwise, run
git rev-parse --show-toplevelto get the current repo root
- If
-
Read
<repo>/.claude/rules/learnings.md(if it exists) -
If the file doesn't exist, create it:
mkdir -p <repo>/.claude/rulesThen write the file with:
# Learnings - <the learning text> -
If the file exists, append
- <the learning text>on a new line at the end -
Confirm:
Added learning to <repo>/.claude/rules/learnings.md
Then stop. Manual add does not run extraction or review.
Extraction Mode
Step 1: Fetch closed, unreviewed tasks
curl -s http://localhost:7777/api/tasks | jq .
Parse the JSON response. Filter to tasks where:
status === "closed"- If a specific
<task-id>argument was provided, also filter to that task
For each closed task, check if it has already been reviewed:
- If
task.worktreeis null or the worktree directory doesn't exist on disk, skip this task silently (worktree was deleted, nothing to review) - The task's
worktreefield gives the worktree path (e.g.,/Users/alice/repos/myapp/.worktrees/abc123) - Check if
<worktree>/.learnings-reviewedexists:stat <worktree>/.learnings-reviewed - If the marker file exists, compare its mtime against the task's
updated_at:- If
updated_atis AFTER the marker's mtime → treat as unreviewed (task was resumed) - If
updated_atis BEFORE or equal → already reviewed, skip
- If
- If the marker file doesn't exist → unreviewed
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.
- 12d ago First seen · 264 lines · 41 tokens per session scan B 1b5718803813
review-learnings is a skill published in the GitHub repository ShreyPaharia/octomux (22 stars, last pushed 11d ago), licensed MIT. It adds 41 tokens to every session and 2,275 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it B with 2 findings (reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
code-review
AI code review for PR or local changes.
change-review
Validate CRM/PM changes before PR.
code-review-quality
Conduct context-driven code reviews focusing on quality, testability, and maintainability. Use when reviewing code, providing feedback, or establishing review practices.
agentplane-task-closure-recovery
Use when Agentplane task completion, direct finish, branchpr integration, hosted-close, close-tail PRs, PR metadata, dirty task artifacts, or remote branch divergence need diagnosis or recovery.
reviewing-code-quality
Reviews a diff or module for slipping standards, favoring deletion over rearranging, and ends in one honest verdict. Use when a change risks oversized files, needless layers, feature logic leaking into shared code, or clever indirection. Do not use for a trivial obvious edit, or for a check that is only about whether…
audit-code
Run a two-pass, multidisciplinary code audit led by a tie-breaker lead, combining security, performance, UX, DX, and edge-case analysis into one prioritized report with concrete fixes. Use when the user asks to audit code, perform a deep review, stress-test a codebase, or produce a risk-ranked remediation plan across…