Borrowing it
Nothing to install: this file belongs to vlad-ryzhkov/ai-context-engineering-for-qa. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/vlad-ryzhkov/ai-context-engineering-for-qa/main/.claude/skills/improve-project-context/SKILL.mdgit 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/improve-project-context)<a href="https://agentmods.dev/skills/vlad-ryzhkov/ai-context-engineering-for-qa/improve-project-context"><img src="https://agentmods.dev/badge/skills/vlad-ryzhkov/ai-context-engineering-for-qa/improve-project-context/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/improve-project-context"><img src="https://agentmods.dev/badge/skills/vlad-ryzhkov/ai-context-engineering-for-qa/improve-project-context.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 | $0.00094 | $0.02517 |
| Opus 5.5 | $0.00038 | $0.01007 |
| Sonnet 5.5 | $0.00019 | $0.00503 |
| Haiku 4.5 | $0.00009 | $0.00252 |
Grade A, and why
improve-project-context 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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/improve-project-context — Cold Topology Audit
ROLE: static analyser. Builds dep graph, measures token budget, finds structural problems. NO session signals, NO JSONL mining (→ improve-context-from-sessions).
Invariants (BANNED to violate)
- Surgical Edit ONLY. Existing files →
Editanchored on header or last 2 lines. New files →Write. Wholesale rewrite = BANNED. - ONE pool, ONE apply. Collect EVERY accepted finding, then apply and commit them together. BANNED: apply a batch → keep auditing → apply again. Each apply re-injects the whole corpus at the next context refresh (a five-figure token floor on a mid-sized corpus), so N applies cost N × that. Re-run
python3 scripts/context-check.pyAFTER the pool, before the commit. - Terse rule style for
rules/**+SKILL.md+ JIT pointers ONLY.docs/**= human-prose OK (rationale, examples, incidents, decision trees). Style spec:docs/context-budget.md. - BANNED reads/edits:
*.local.json, gitignored paths (verify FIRST),*.zip/binary (flag ORPHAN-ARTIFACT). - Trigger: explicit user call ONLY (
/improve-project-context, "audit context", "cold audit"). NO auto-fire. - PLACEMENT: rule unique-to-project AND useful-to-colleagues →
<repo>/.claude/(git-shared). Global rule OR personal preference →~/.claude/(CLAUDE.md/rules/docs). - BANNED project→global ref: shared project context (repo CLAUDE.md /
.claude// docs/) MUST NOT depend on~/.claude/**— dangling in a colleague's clone. Inline the gist or keep it in repo.claude/. A~/.clauderef is allowed ONLY as an explicitly-optional personal add-on. FLAG every such ref in audits. - BANNED hardcoded feature-branch in shared files: scripts / docs / templates committed for colleagues MUST NOT pin a personal branch (
--ref feat/...,git checkout feat/...). Usemainor a<your-branch>placeholder. FLAG every committed--ref <non-main-branch>/ branch literal in audits. - SHRINK SKILL.md over 300: TWO levers, in order — (1) EXTRACT blocks (examples, algorithms, tables, rationale) →
references/(JIT, 0 start-cost); (2) THEN compress remaining prose perdocs/terse-rule-style.md. Extract beats compress — extracted = not loaded at all; compressed = still loaded, just smaller. - COMPRESS prose-heavy CLAUDE.md / rules/*.md over budget → RECOMMEND a compress pass per
docs/terse-rule-style.md. Other.md→ softer recommend, only if token-pressured. - YAGNI gate: corpus < ~5 context files (1 CLAUDE.md + few configs) → SKIP context-map + SDLC-phase ceremony. Bureaucracy beyond value. Run full pipeline only when files are pile-up-ing or multi-agent.
- COLD blindness: this skill is STATIC (file topology). It can't see how the model BEHAVES (a rule the AI keeps violating in dialog passes a cold audit). Pair with warm audit →
improve-context-from-sessions. State this when reporting: "static-clean ≠ behaviorally-followed". - DOGFOOD: this skill's own SKILL.md + references ARE valid audit targets. Run on self/siblings periodically — caught real orphans + AUTO-conflicts in practice.
What ships with it
9 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.
- references/audit-modes.md 10 KB
- references/context-map-format.md 3.0 KB
- references/context-rubric.md 4.7 KB
- references/dependency-graph-algorithm.md 6.9 KB
- references/infrastructure-layers.md 8.1 KB
- references/output-formatting.md 8.1 KB
- references/patch-card-example.md 2.1 KB
- references/sdlc-map-workflow.md 3.3 KB
- references/skills-vs-agents.md 2.6 KB
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 First seen · 91 lines · 94 tokens per session scan A 9918a108624d
improve-project-context 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 94 tokens to every session and 2,517 once invoked, about $0.0004 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 2026-09-09.
Other skills, from other repositories
effective-kotlin
Apply Effective Kotlin best practices (Marcin Moskała, 2nd Ed). Covers Safety (Items 1-10: mutability, scope, nulls, types, expectations, errors, resources, tests), Readability (Items 11-18: operators, receivers, properties, naming), Reusability (Items 19-25: DRY, generics, delegation, variance), Abstraction (Items…
kotlin-in-action
Apply Kotlin In Action practices (Elizarov, Isakova, Aigner, Jemerov, 2nd Ed). Covers Basics (Ch 1-3: functions, extensions, default args), Classes (Ch 4: sealed, data, delegation, companion objects), Lambdas (Ch 5-6: functional APIs, sequences, scope functions), Nullability (Ch 7-8: safe calls, Elvis, platform types…
openspec-new-change
Start a new OpenSpec change using the experimental artifact workflow. Use when the user wants to create a new feature, fix, or modification with a structured step-by-step approach. Also use when the user says "openspec new change" or "opsx new".
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
memory-system
A guide to giving an AI agent persistent, user-specific memory across conversations. It covers what to remember, how to store it, how to update or compress it, and when to use memory instead of a shared knowledge base.
writing
A writing guide for turning verified facts and calculations into finished text for a specific audience. It follows the requested language, structure, and length.