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 mattgierhart/PRD-driven-context-engineering --skill prd-v08-drift-baseline-comparegit clone --depth 1 https://github.com/mattgierhart/PRD-driven-context-engineeringWrote 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/mattgierhart/prd-driven-context-engineering/prd-v08-drift-baseline-compare)<a href="https://agentmods.dev/skills/mattgierhart/prd-driven-context-engineering/prd-v08-drift-baseline-compare"><img src="https://agentmods.dev/badge/skills/mattgierhart/prd-driven-context-engineering/prd-v08-drift-baseline-compare/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/mattgierhart/prd-driven-context-engineering/prd-v08-drift-baseline-compare"><img src="https://agentmods.dev/badge/skills/mattgierhart/prd-driven-context-engineering/prd-v08-drift-baseline-compare.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 | $0.00111 | $0.02540 |
| Opus 5 | $0.00056 | $0.01270 |
| Sonnet 5 | $0.00022 | $0.00508 |
| Haiku 4.5 | $0.00011 | $0.00254 |
Grade A, and why
prd-v08-drift-baseline-compare 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 11d 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 — 214 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Drift: Baseline / Compare / History
Position in workflow: v0.8 Monitoring Setup → v0.8 Drift: Baseline / Compare / History → v0.8 Runbook Creation
Execution Mode
Default is standard. See .claude/rules/08-skill-execution-modes.md for selection logic.
| Mode | What this skill produces |
|---|---|
| quick | One metric / config / dataset baselined; weekly compare schedule; simple threshold alert |
| standard | 3–5 things monitored; baseline + tiered thresholds (warn / critical); compare cadence; history retention |
| deep | Full portfolio; multi-dimensional comparison (per segment, per environment); regression-cause matrix; auto-baselining after intentional change |
What This Does
Generalizes a pattern AgriciDaniel's claude-seo skill encodes for SEO drift — baseline → snapshot → compare → history — into a reusable monitoring shape that works for any metric, config, or dataset that can change over time and needs to be watched.
This is drift monitoring, distinct from alerting on absolute thresholds. Alerting answers "is X over the line right now?" Drift monitoring answers "is X different from last week's normal?" — which catches slow regressions that absolute thresholds miss.
Examples of things worth drift-monitoring:
- KPI: activation rate week-over-week
- AI search position: ChatGPT/Perplexity ranking for target queries
- Config: feature-flag rollout percentages
- Performance: p95 latency by endpoint
- Cost: per-user infra cost
- Marketing: per-channel CAC trend
- Content: changelog post engagement
- Third-party: vendor pricing pages (price hikes), competitor feature pages (parity loss)
How It Works
- Pick what to monitor — One thing per MON-DRIFT- entry. Must be:
- Quantifiable (number, percentage, list, configuration value)
- Snapshotable (captureable at a point in time, ideally automatically)
- Causally interpretable (when it changes, you know enough to investigate)
- Capture the baseline — Take a snapshot. Date it. Store in version control or a known location (
status/baselines/,monitoring/snapshots/, etc.). - Define drift thresholds:
- Warn: meaningful change (e.g., 10% drift in a KPI; any change in a config value)
- Critical: serious change (e.g., 25% KPI drop; breaking config change)
- Recalibrate: intentional change that should refresh the baseline (e.g., after a feature rollout, the baseline is wrong; refresh it)
- Set compare cadence — How often does this get re-snapshotted?
- Hot (hourly/daily): production KPIs, AI search positions during a launch
- Warm (weekly): standard product KPIs, content engagement
- Cool (monthly): vendor pricing, competitor feature parity, infra cost
- Build the compare procedure — A script or runbook that:
- Takes a new snapshot
- Diffs against baseline
- Computes drift % per dimension
- Emits warn/critical signals at thresholds
- Appends to history log
- Plan auto-baselining after intentional change [standard+] — When the team makes a deliberate change (ships a feature that should improve activation), the old baseline becomes wrong. Define what triggers a baseline refresh and who approves it.
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.
- 11d ago First seen · 214 lines · 111 tokens per session scan A afdbb7746cfc
prd-v08-drift-baseline-compare is a skill published in the GitHub repository mattgierhart/PRD-driven-context-engineering (182 stars, last pushed 10d ago), licensed MIT. It adds 111 tokens to every session and 2,540 once invoked, about $0.0006 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
ralplan
Consensus planning entrypoint that auto-gates vague ralph/autopilot/team requests before execution.
remember
Review reusable project knowledge and decide what belongs in project memory, notepad, or durable docs.
embedding-strategies
Select and optimize embedding models for semantic search and RAG applications. Use when choosing embedding models, implementing chunking strategies, or optimizing embedding quality for specific domains.
docx-comment-reply
Reply to comments (批注) in Word .docx/.doc files: extract comment context, draft replies, write threaded replies back, and validate OOXML.
explaining-machine-learning-models
Explain trained machine learning models through feature attribution, local explanations, and behavior summaries. Use as an explicit/manual helper once a model already exists, not for training ownership, leakage auditing, or general ML strategy selection.
aiwg-regenerate-copilot
Regenerate copilot-instructions.md for GitHub Copilot with vendor-specific content only.