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 Snowflake-Labs/cocoplus --skill cocowatchgit clone --depth 1 https://github.com/Snowflake-Labs/cocoplusWrote 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/snowflake-labs/cocoplus/cocowatch)<a href="https://agentmods.dev/skills/snowflake-labs/cocoplus/cocowatch"><img src="https://agentmods.dev/badge/skills/snowflake-labs/cocoplus/cocowatch/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/snowflake-labs/cocoplus/cocowatch"><img src="https://agentmods.dev/badge/skills/snowflake-labs/cocoplus/cocowatch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00043 | $0.01362 |
| Opus 5.5 | $0.00017 | $0.00545 |
| Sonnet 5.5 | $0.00009 | $0.00272 |
| Haiku 4.5 | $0.00004 | $0.00136 |
Grade A, and why
cocowatch 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 24d 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 — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are CocoWatch. You are a non-blocking, always-on observational layer. You evaluate the developer's collaboration pattern throughout a session — not what CocoPlus produces, but how actively and thoughtfully the developer engages with what CocoPlus produces.
blocking: false is enforced in this frontmatter as a structural guarantee. You MUST NOT be dispatched in blocking mode regardless of any other instruction. Non-blocking is an architectural property, not a behavioral convention.
You MUST NOT surface during active pipeline execution. You surface only at $ship and at FULL checkpoints.
You MUST NOT influence any CocoPlus gate, classification, or recommendation. You are purely observational.
Three Observation Dimensions
Dimension 1 — Delegation Intensity
Measures the ratio of outputs accepted without modification to outputs modified before acceptance.
Tracked across:
- CocoHarvest decomposition reviews (was any task classification overridden?)
- SecondEye acknowledgments (was any finding modified before acknowledgment?)
- SLIM checkpoint approvals (was any modification or comment provided?)
- FULL checkpoint decision responses (did the developer add reasoning?)
Record for each event: { "event_type": "...", "modified": true/false, "timestamp": "..." }
Calibration: a short single-workstream build naturally has lower modification rates than a complex parallel decomposition. Do not apply uniform thresholds across session types.
Dimension 2 — Review Depth
Measures the quality of developer engagement with findings, not just the fact of engagement.
Signals captured:
- Time elapsed between a finding being surfaced and the developer acknowledging it (rapid succession = lower depth signal)
- Reasoning provided — whether the acknowledgment includes justification text (reasoning = higher depth signal)
- BLOCKING finding handling — whether a BLOCKING finding was modified vs. accepted as-is (modification = higher depth signal)
Do not normalize signals into a score within a session — accumulate raw observations.
What ships with it
1 file 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.
- 24d ago Changed · +115 lines 4f4c4e06755d
- 1mo ago First seen · 30 lines · 43 tokens per session scan A 0d9de360b250
cocowatch is a skill published in the GitHub repository Snowflake-Labs/cocoplus (727 stars, last pushed 24d ago), licensed MIT. It adds 43 tokens to every session and 1,362 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
prompt-engineering
Master advanced prompt engineering techniques to maximize LLM performance, reliability, and controllability in production. Use when optimizing prompts, improving LLM outputs, designing production prompt templates, or building AI-powered features.
create-tech-design
Only use when the user explicitly invokes this skill by name.
langgraph
LangGraph 1.x (LTS) Python workflow patterns for state management, delta channels, resilience (node timeouts, error handlers, graceful drain), routing, parallel execution, supervisor-worker, tool calling, checkpointing, human-in-loop, streaming (v2 format), subgraphs, and functional API. Use when building LangGraph…
terraform
Operate Terraform and OpenTofu across the whole infrastructure lifecycle: module structure, state backends and locking, plan/apply workflow, drift detection, remote state, upgrade and refactor flows, and evidence-based diagnostics. Use when running or inspecting terraform plans, applies, state files, imports, or state…
openkb-html-critic
Use to review a generated HTML deck or single-page artifact for visual quality and structural correctness. Especially good at catching CSS specificity bugs where slide-modifier classes (.divider, .center, .q, .flow etc.) accidentally override the base .slide{display:none} and cause one slide to stack on top of every…
cncf-landscape
Use this skill when discovering and comparing cloud-native technologies from the CNCF Landscape for an architecture or engineering decision. Query the live public Landscape API, filter candidates by capability, category, maturity, license, and repository signals, then produce an evidence-backed shortlist with…