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 skills add shinpr/codex-workflows --skill recipe-outcome-watchgit clone --depth 1 https://github.com/shinpr/codex-workflowsWrote 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/shinpr/codex-workflows/recipe-outcome-watch)<a href="https://agentmods.dev/skills/shinpr/codex-workflows/recipe-outcome-watch"><img src="https://agentmods.dev/badge/skills/shinpr/codex-workflows/recipe-outcome-watch/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/shinpr/codex-workflows/recipe-outcome-watch"><img src="https://agentmods.dev/badge/skills/shinpr/codex-workflows/recipe-outcome-watch.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 | $0.00033 | $0.00603 |
| Opus 5.5 | $0.00013 | $0.00241 |
| Sonnet 5.5 | $0.00007 | $0.00121 |
| Haiku 4.5 | $0.00003 | $0.00060 |
Grade A, and why
recipe-outcome-watch 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 4d 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.
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 — 33 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Outcome Watch
Run a finite, independent direction review alongside the current task. Use governing documents or an explicit purpose, with any development task.
Request: $ARGUMENTS
Resolve the basis
- Use governing document paths or an explicit purpose supplied in the arguments. Pass documents as paths, preserving their authority.
- Otherwise inspect documents under the working directory and their relationship to the current task. Select a document whose content establishes that relationship, with its relevant references. If multiple plausible targets remain, ask the user to select one.
- When no governing document exists, use the user's explicit task purpose. If you must infer or synthesize the purpose or completion condition from the session, present that wording for confirmation before starting and pass the confirmed wording unchanged.
Reuse existing task evidence. Supply an exact activity-log path when available and relevant. Resolve a changed monitoring scope from explicit user instructions and restart with that basis.
Start and deliver
Resolve scripts/watch.py relative to this skill and use an available Python 3 interpreter. The script owns timing, codex exec reviews, and cleanup. Run it through the host's asynchronous execution facility, forwarding its output automatically into the active session while continuing the task. Establish that delivery from the available tool contracts before starting. If the host offers only manual polling or next-turn message queues, explain that live delivery is unavailable and end this recipe.
Start in the target directory with --source PATH for each governing document, or --purpose TEXT for the explicit or confirmed purpose. Optional --activity PATH supplies existing evidence; --cwd PATH selects the target directory. Quote arguments for the available shell. Use --interval SECONDS, --duration SECONDS, and --language LANGUAGE for the user's choices.
Defaults are a 60-minute interval and a 4-hour duration. The first review runs immediately, with subsequent starts spaced by the interval. Each review is bounded by 10 minutes and the remaining duration. Use the host's normal permission mechanism to launch the authenticated CLI; reviews run read-only.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 4d ago First seen · 33 lines · 33 tokens per session scan A ff7446a1d076
recipe-outcome-watch is a skill published in the GitHub repository shinpr/codex-workflows (39 stars, last pushed 3d ago), licensed MIT. It adds 33 tokens to every session and 603 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-10-12.
Other skills, from other repositories
linear-tickets
Create or update high-quality Linear tickets while preserving the originating report, diagnostic links, attachments, and reporter/requester attribution. Read this whenever someone asks to create, recreate, or update a Linear issue.
run-task
Execute a single Todo task through In Progress to Review, meeting every acceptance criterion with tests and vibe-lint checks. Refuses Planning-status tasks. Invoked as /agiflow:run-task . Uses gettask, updatetask, createtaskcomment.
chrome-cdp
Drive a headless Chrome over the Chrome DevTools Protocol (CDP) for browser QA — navigate, click, fill forms, read the DOM/accessibility tree, screenshot, and assert. Use whenever a task requires loading a web page and interacting with it like a user. Chrome is launched by a bash step (recipe below); this skill…
agent-phase-closeout
A Chinese-language process for closing out a completed work stage and recording its state. It summarizes finished work, preserves decisions, freezes completed interfaces, and writes a resume checkpoint.
linear-app-setup
Create and configure a Linear agent app so the assistant can manage issues, projects, and workflows under its own identity.
blueprint
Turn a one-line objective into a step-by-step construction plan for multi-session, multi-agent engineering projects. Each step has a self-contained context brief so a fresh agent can execute it cold. Includes adversarial review gate, dependency graph, parallel step detection, anti-pattern catalog, and plan mutation…