v0.3.4 - context cost estimator

v0.3.4 - context cost estimator

skill-doctor v0.3.4 adds a local context-cost check for AI agent instruction layers.

Highlights

• Adds skill-doctor cost and skill-doctor context as aliases for estimating per-turn context token tax.
• Grades the estimate against a configurable token budget.
• Adds --fail-on-budget so teams can use the check in CI before always-on instructions grow too large.
• Adds JSON output for automation and dashboards.

Why it matters

Agent setups now often include Claude skills, Cursor rules, Copilot instructions, Codex AGENTS.md, and other always-on files. Even when those files are safe, they can quietly add context cost to every turn.

The new command gives users a concrete number and a short recommendation for the highest-cost items.

Try it

npx @evilstar2025/skill-doctor cost
npx @evilstar2025/skill-doctor cost --budget-tokens 2000 --fail-on-budget
npx @evilstar2025/skill-doctor context --json

Safe demo:

git clone https://github.com/evilstar2016/skill-doctor.git
cd skill-doctor/examples/conflicted-agent-project
npx @evilstar2025/skill-doctor scan --scope project
npx @evilstar2025/skill-doctor conflicts --scope project
npx @evilstar2025/skill-doctor audit --scope project
npx @evilstar2025/skill-doctor cost --scope project

Feedback wanted

I am looking for real, redacted examples of:

• always-on agent instruction files that feel too expensive for every turn
• false positives or misleading recommendations from skill-doctor cost
• missing agent platform paths or instruction file patterns
• cases where overlapping AI agent rules made a tool behave inconsistently

Actionable reports: #4

Lightweight questions and examples: #6

The project reads local files and reports findings. It does not upload your skill files by default.

Verification

• npm run build passed.
• npm test passed: 39 test files, 329 tests.
• Official npm latest is @evilstar2025/skill-doctor@0.3.4.