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 vlad-ryzhkov/ai-context-engineering-for-qa --skill curate-lessonsgit clone --depth 1 https://github.com/vlad-ryzhkov/ai-context-engineering-for-qaWrote 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/vlad-ryzhkov/ai-context-engineering-for-qa/curate-lessons)<a href="https://agentmods.dev/skills/vlad-ryzhkov/ai-context-engineering-for-qa/curate-lessons"><img src="https://agentmods.dev/badge/skills/vlad-ryzhkov/ai-context-engineering-for-qa/curate-lessons/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/vlad-ryzhkov/ai-context-engineering-for-qa/curate-lessons"><img src="https://agentmods.dev/badge/skills/vlad-ryzhkov/ai-context-engineering-for-qa/curate-lessons.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 | $0.00045 | $0.01521 |
| Opus 5.5 | $0.00018 | $0.00608 |
| Sonnet 5.5 | $0.00009 | $0.00304 |
| Haiku 4.5 | $0.00005 | $0.00152 |
Grade A, and why
curate-lessons 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 1mo 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 — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/curate-lessons — Lesson Curation
Curates lessons from .ai-lessons/pending.md, deduplicates against existing rules, and promotes
confirmed patterns into target context files.
When to Use
Run when .ai-lessons/pending.md has >= 3 entries.
Phase Checkpoints
STOP if:
.ai-lessons/pending.mddoes not exist → outputcurate-lessons: .ai-lessons/pending.md not found. Create it first.and STOP- Fewer than 3 entries in
pending.md→ output count warning and STOP
WARN if:
- A rule has no
Source:orDate:metadata → tag as unverified, proceed with caution
INFORM:
- Phase transitions only (no per-rule progress lines)
SILENT MODE
Output only phase transitions and the final SKILL COMPLETE block. No intermediate progress lines.
Phase 1: Load Pending Lessons
Read .ai-lessons/pending.md. Extract all RULE: entries with source and date metadata.
Entries may have a [REFLECTOR] prefix — these come from the Reflector protocol (proactive pattern detection). Preserve the pattern metadata (Pattern: {detector}, {count}x across {skills}) in the curation report for traceability. [REFLECTOR] entries have already been semantically deduped by the Reflector protocol, but Phase 2 still runs its own dedup pass (defense in depth).
If file is empty or has < 3 entries → output:
curate-lessons: only {N} pending entries — minimum 3 required. Run again when more lessons accumulate.
and STOP.
Phase 2: 2-Pass Hybrid Dedup
Pass 1 — Grep Narrowing: For each pending rule, extract 2-3 keywords. Run Grep across dedup targets:
CLAUDE.md.claude/qa-antipatterns/**/*.md.claude/skills/*/SKILL.md.claude/protocols/*.md
Collect only files with keyword hits → candidate set.
Pass 2 — Semantic Comparison (candidates only): Read each candidate file. Compare rule intent semantically.
Assign one of 4 verdicts:
| Verdict | Meaning | Action |
|---|---|---|
DUPLICATE |
Identical or semantically equivalent rule already exists | Skip — do not promote |
OVERLAP |
Partial overlap with existing rule | Flag for user — suggest merging |
UNIQUE |
No coverage in any target file | Promote |
META |
Self-referential: about curation/gardener process | Target: curate-lessons/SKILL.md or protocols/gardener.md, NOT CLAUDE.md |
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.
- 1mo ago Changed · +141 lines · +24 tokens per session 5089a1a34aed
- 1mo ago First seen · 25 lines · 21 tokens per session scan A 5c819a785990
curate-lessons is a skill published in the GitHub repository vlad-ryzhkov/ai-context-engineering-for-qa (6 stars, last pushed 1mo ago), licensed Unlicense. It adds 45 tokens to every session and 1,521 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 2026-08-31.
Other skills, from other repositories
agent-state-bar
A design guide for an agent status bar: a short, automatically updated summary of progress, remaining tasks, tool counts, errors, and system state placed near the latest model input.
gsd-graphify
Build, query, and inspect the project knowledge graph in .planning/graphs/.
context-compression
A guide to shortening the information an agent carries between steps. It explains how to replace large tool outputs and old conversation details with compact, useful summaries while preserving important decisions and constraints.
handoff
Prepare a clean cross-session handoff so the next agent can pick up where you left off. Saves a focused Work Checkpoint with gsdcheckpointsave and ensures summary artifacts are current. Use when asked to "hand off", "prepare handoff", "pause work", "bookmark this", "I'll come back to this later", before running out of…
02-project-memory
Build the project's memory of its architecture, conventions, and decisions, and wire it into your AI tools. Use when the user wants to set up or refresh project memory, or rewire it into a tool. Not for editing one existing memory file.
learn
Diagnose and fix agent behavioral surfaces when the user corrects a mistake — connects to Claude native memory.