improve allure agent mode

Author: baev

AGENTS.md

@@ -1,14 +1,12 @@
 # Project Guide
 
-Use [Allure Agent Mode](docs/allure-agent-mode.md) for all test-related work in this repository.
+Use [Allure Test Agent](docs/allure-test-agent.md) for all test-related work in this repository.
 
-- Read `docs/allure-agent-mode.md` before designing, writing, reviewing, validating, debugging, or enriching tests.
-- If a command executes tests and its result will be used for smoke checking, reasoning, review, coverage analysis, debugging, or any user-facing conclusion, run it through `allure agent`. It preserves the original console logs and adds agent-mode artifacts without inheriting the normal report or export plugins from the project config.
-- Use `allure agent` for smoke checks too, even when the change is small or mechanical.
-- After changing a package in this repository, treat package builds as part of validation and run the changed package build command before finalizing (for example, `yarn workspace <package-name> build`), because type-level regressions can break builds.
-- **Workspace build order:** dependency edges come from each package's `dependencies` (including `workspace:*` links to other packages). The root `yarn build` script runs `yarn workspaces foreach -Avvpt run build`, where `-t`/`--topological` schedules a workspace only after its regular `dependencies` have built successfully (parallelism `-p` is within those constraints). For a single touched package, `yarn workspace <package-name> build` is enough; when several packages change or downstream breakage is possible, prefer a full **`yarn build`** at the repo root so Yarn applies the same topological order as CI.
-- **Lint, format, and type-aware lint:** before finalizing code changes, align with the `lint` job in `.github/workflows/build.yml`: run **`yarn build`**, then **`yarn lint`**, **`yarn format:check`**, and **`yarn lint:type`**. The root script `yarn verify` runs `format:check`, `lint`, and `lint:type` together but does not run `yarn build`; use it only when you already built and need the same static checks in one command.
+- Read `docs/allure-test-agent.md` before designing, writing, reviewing, validating, debugging, or enriching tests. Use the `$allure-test-agent` skill as the durable behavior guide when it is installed; this project file contains local commands and conventions.
+- If a command executes tests and its result will be used for smoke checking, reasoning, review, coverage analysis, debugging, or any user-facing conclusion, run it through the local agent test service when available, or through `allure agent` otherwise. It preserves the original console logs and adds agent-mode artifacts without inheriting the normal report or export plugins from the project config.
+- Use agent-mode execution for smoke checks too, even when the change is small or mechanical.
 - Only skip agent mode when it is impossible or when you are debugging agent mode itself.
-- If agent-mode output is missing or incomplete, debug that first rather than silently falling back to console-only review; use the checklist in `docs/allure-agent-mode.md` (**Agent mode failures and unavailable runs**).
+- If agent-mode output is missing or incomplete, debug that first rather than silently falling back to console-only review.
 - Use Allure agent-mode when adding tests for features or fixes so expectations, evidence quality, and scope review are part of the loop.
-- Keep any non-testing project guidance here short; the detailed Allure workflow belongs in the linked guide.
+- Do not present ignored, excluded, swallowed, or non-gating tests as a passing validation signal.
+- After changing a package, run the changed package build command before finalizing. For several packages or possible downstream breakage, prefer the root `yarn build` so Yarn applies the same topological order as CI.

README.md

@@ -74,7 +74,13 @@ For example:
 npx allure agent -- npm test
 ```
 
-`allure agent` runs with an agent-only profile by default. It creates a fresh output directory automatically, can load an expectations file with `--expectations`, and ignores configured presentation or export plugins such as Awesome or TestOps unless you explicitly fall back to the lower-level `ALLURE_AGENT_*` plus `allure run` flow.
+`allure agent` runs with an agent-only profile by default. It creates a fresh output directory automatically, accepts compact inline expectations such as `--goal`, `--expect-tests`, `--expect-test`, `--expect-label`, and `--expect-step-containing`, can load an expectations file with `--expectations`, and ignores configured presentation or export plugins such as Awesome or TestOps for that run.
+
+Agents and setup tools can inspect the local structured capability contract without scraping help text:
+
+```bash
+npx allure agent capabilities --json
+```
 
 ### Generating Reports Manually
 
@@ -122,6 +128,7 @@ The Allure CLI includes several helpful global options. Use `--help` to explore
 
 ```bash
 npx allure run --help
+npx allure agent capabilities --json
 npx allure agent --help
 npx allure watch --help
 ```