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 skills/etr/groundwork/staged-rolloutnpx skills add etr/groundwork --skill staged-rolloutgit clone --depth 1 https://github.com/etr/groundworkWhat 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 | $0.00041 | $0.00910 |
| Opus 5 | $0.00020 | $0.00455 |
| Sonnet 5 | $0.00008 | $0.00182 |
| Haiku 4.5 | $0.00004 | $0.00091 |
Grade A, and why
staged-rollout 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 2d 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 — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Staged Rollout
Overview
A change reaches all users gradually, behind a switch you can flip back, while someone watches the signals. Big-bang releases convert a small bug into a full outage. A staged rollout converts the same bug into a contained blip caught at 1% traffic.
Core principle: Every outward-facing change ships behind a flag, advances in stages with a monitoring window between them, and has a rollback plan written before launch — not improvised during the incident.
This depends on the change being observable first: see [[instrument-observability]] for the signals the monitoring window watches.
When to Use
- Any user-facing or externally-observable change (new endpoint, UI, behavior change, schema migration)
- Risky internal changes where a bad deploy degrades many users at once
Skip only for trivially reversible, low-blast-radius changes (a copy fix, an internal doc). When unsure, stage it — the cost is one flag.
Process
-
Pre-launch checklist. Confirm before any traffic shift:
- Validation and review already passed (see [[validate]] / [[review-pr]])
- Instrumentation is live and the dashboards read real data
- The feature flag exists and defaults to off
- The rollback plan is written (step 4)
-
Feature-flag gating. Put the new behavior behind a flag, default off. The flag must be flippable at runtime without a redeploy — that is what makes rollback fast.
-
Staged / canary rollout. Advance through stages, not in one jump. A typical ramp:
Stage Audience Hold for Canary 1% (or internal/dogfood) a monitoring window Early 10% a monitoring window Majority 50% a monitoring window Full 100% — Monitoring window: between each stage, watch the RED metrics and alerts long enough to span real traffic before promoting. Do not promote on a clean dashboard you've watched for thirty seconds. If a symptom breaches, stop and roll back — do not push forward hoping it settles.
-
Written rollback plan. Before launch, document: the exact trigger conditions (which metric/alert at which threshold), the precise rollback action (flip flag
Xto off; revert migrationY), who can execute it, and the expected recovery time. A rollback that requires a redeploy is too slow — prefer the flag.
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.
- 2d ago First seen · 68 lines · 41 tokens per session scan A 8ed2adb564a2
staged-rollout is a skill published in the GitHub repository etr/groundwork (42 stars, last pushed 21d ago), licensed MIT. It adds 41 tokens to every session and 910 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 skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
agent-host-chat-contributions
Build and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.