49 lines (34 loc) · 1.36 KB

AGENTS.md

lintlang is a static linter for AI agent configs, tool definitions, and system prompts.

Use it for

linting tool descriptions before agents start choosing the wrong tools
checking prompts and configs for missing constraints, schema mismatches, and role confusion
running a zero-LLM CI gate over YAML, JSON, and prompt text

Do not use it for

runtime evaluation
dynamic agent testing
proving an agent is safe in production

Minimal commands

pip install -e ".[dev]"
lintlang --help
lintlang scan samples/bad_tool_descriptions.yaml
pytest -q
ruff check src/ tests/

Output shape

terminal verdicts: PASS, REVIEW, or FAIL
structural findings by pattern H1 through H7
JSON output for CI via --format json

Success means

the same config file produces the same verdict and findings
scan output points to concrete locations and rewrite guidance
tests and sample self-scan pass offline

Common failure cases

users expect lintlang to judge runtime model behavior
configs are syntactically valid but still too underspecified to be safe
teams gate only on syntax linters and miss language-level failure modes

Maintainer notes

keep detector language aligned with actual failure modes in the samples
keep CLI examples and severity semantics aligned with README
keep the tool fully offline and deterministic