Quality is a skill-only plugin that helps coding agents operate project-owned quality capabilities. It does not ship test suites, evaluation datasets, or business judgment criteria, and it does not replace the frameworks and project assets that make verification executable.
The bundled skills currently cover three independent quality views:
e2eoperates a project's existing end-to-end capability and incrementally adds project-owned coverage when requested;perfoperates and compares project-owned load, capacity, stress, and soak experiments, and can incrementally add bounded performance coverage when requested;trajectoryoffers Job, Session and Context views to connect results, execution and prompt organization, assess Good/Fast/Cheap (effect, speed and cost), and design optimization experiments.
The project owns the tests, profiles, recordings, adapters, and judgment criteria. A framework such as quality-harness supplies execution and analysis mechanisms; these skills discover and operate the project's canonical entrypoints, interpret evidence, and preserve useful operating knowledge beside the project assets.
The skills share Quality concepts: subject and environment identity, execution facts, evidence and evaluation, comparison, and the distinction between capability, execution health, and quality decisions. Each skill retains its domain's lifecycle and judgment rules.
identify the question and subject
→ discover a project capability or sufficient recorded evidence
→ prepare live execution or select existing evidence
→ use the canonical entrypoint
→ interpret results and coverage
→ report the conclusion and next supported action
Live E2E and performance runs follow environment preparation: select
the target and revision, locate the test process, verify application connectivity, prepare
conditions, and retain cleanup evidence. Offline analysis reuses recorded facts without requiring
a live target. No usable capability or sufficient evidence is reported as no_capability;
an unavailable prerequisite is blocked.
Execution facts and judgment remain separate. A passed assertion or SLO covers only its realized conditions; missing observations, skipped checks, and execution errors remain visible. Skills can propose the next experiment or improve project-owned assets when requested. They neither invent a missing suite nor supply cross-skill execution orchestration.
Job View, Session View and Context View are peer analysis entry points. Job View follows a business goal and its results; Session View follows agent execution; Context View makes prompt organization readable by actual message role and internal sections. Recorded identities connect results to decisions and the prompts behind them. Use the Agent Good–Fast–Cheap triangle to assess effect, speed and cost together, then use the views to explain the observed behavior.
In conversation or TUI, asking to see a SessionView produces a text presentation of actual model requests, responses and tool calls, making context organization and prompt issues inspectable. “Loop view” and “agent view” are accepted request phrases for that presentation; HTML is another rendering of the same observation.
Provide a project entrypoint, saved observation or existing evaluation artifact. The skill establishes scope and available goals, annotations and generation configuration before diagnosing failures and avoidable work. Shared calls retain their identity across views so task associations do not duplicate measured cost. See Observation, Evidence and Evaluation.
Add the devloop marketplace once, then install the plugin.
Claude Code:
/plugin marketplace add https://github.com/compforge/devloop.git
/plugin install quality@devloop
Codex:
codex plugin marketplace add https://github.com/compforge/devloop.git
codex plugin add quality@devloopStart a new session after installation so the agent can discover the bundled skill.
Ask for the outcome instead of naming a framework command:
Run this project's E2E tests and report the result.
Check whether the test environment passes E2E.
Find and execute the existing E2E capability; tell me if the project has none.
Add one bounded E2E case for this regression and run the nearest relevant checks.
Run the project's smallest capacity profile and report the trustworthy limit and caveats.
Compare the latest performance run with its declared baseline without generating load again.
Add one bounded soak or scaling profile using the project's existing workload adapter.
Evaluate this project's agent trajectories: how well did it work, how long did it take, and what did it cost?
Find the next trajectory problem worth optimizing.
For this high-token trajectory problem, identify likely causes and controlled experiments.
The skills operate existing project-owned quality assets. Assessment does not invent missing tests, load profiles, recordings, labels, or judgment components. When the user asks to grow coverage, changes remain in the project-owned capability and move toward broader coverage incrementally. The skills do not deploy environments without authorization, modify agent behavior unless asked, or turn unavailable coverage into a pass.
Run make lint test in quality/. Lint checks plugin metadata, skill headers and local documentation
links; tests cover the shipped assets and the validator's failure cases. These checks validate the
plugin's static assets, not model behavior or the quality of reviews on project trajectories.