[OPIK-7167] [QA] feat: dev testing workflow — /explore-feature skill + release-gate scaffolding

[AndreiCautisanu]

Details

Adds the QA-owned /explore-feature skill for the dev-driven testing workflow (automation-coverage initiative): each PR ships one happy-path test that gates its own release; QA reviews and integrates the keepers. This PR is the skill and its scaffolding only.

• /explore-feature skill (.agents/skills/explore-feature/, mirrored to .claude/): a thin orchestrator that resolves scope (local branch diff / a PR / multiple PRs — including "I haven't opened a PR yet"), writes a happy-path test-plan.md to Jira, and delegates the actual test authoring to writing-e2e-tests via a fixed staging-ready contract. The version stamp is read from origin/main:version.txt and encoded as a @release-gate:<version> tag so gate selection is a floor comparison that survives cancelled/skipped releases.
• E2E scaffolding: tests/_release-gate/ + a lifecycle README, and a test:release-gate grep target isolated from the curated @t1/@t2/@t3 tiers; a note in the writing-e2e-tests conventions documents the _release-gate/ exception so those specs don't read as guideline violations.

The skill was hardened by piloting it end-to-end on two real changes (a fix that needed a repro-specific seed, and a no-behavior-delta perf change), which shaped its guardrails: gate the repro not the easy path; seeding support is in-scope authoring; probe backend health via <baseUrl>/api; local-diff mode for pre-PR branches; a perf-fix override for no-behavior-delta changes. The pilot test/bridge artifacts are intentionally NOT part of this PR — they were validation, not deliverables.

Out of scope (deferred follow-up on the ticket): CI wiring — the staging gate selection by stamp, the post-ship move to shipped/, and the age-expiry job.

Change checklist

• User facing
• Documentation update

Issues

• Resolves OPIK-7177

AI-WATERMARK

AI-WATERMARK: yes

• Tools: Claude Code
• Model(s): Claude Opus 4.8
• Scope: full implementation (brainstorm → plan → skill authoring + scaffolding, hardened via two end-to-end pilots), human-directed
• Human verification: design decisions made by the author; skill piloted end-to-end against a live local stack on two real changes (both produced green, relevant release-gate tests during validation — those artifacts are not committed here); grep-target isolation and skill frontmatter verified

Testing

• cd tests_end_to_end/e2e && npm ci, then verified the scaffolding: npm run test:release-gate -- --list selects only @release-gate specs; npm run test:t1/t2/t3 -- --list are unchanged (20/32/36) and the @release-gate tag does not leak into the curated tiers.
• Validated the skill files: SKILL.md frontmatter parses, cross-references resolve, and make claude mirrors cleanly to .claude/.
• During validation (not part of this PR) the skill was driven end-to-end on two real changes and produced green release-gate tests against a live local stack; the #7322 gate's relevance was proven by going RED against the pre-fix code.
• Not run: the full E2E suite against staging (deferred CI step).

Documentation

• tests_end_to_end/e2e/tests/_release-gate/README.md documents the gate-test lifecycle for devs/QA.
• The skill files (SKILL.md, test-plan-template.md, release-gate-contract.md) are the operator-facing documentation; the writing-e2e-tests conventions now note the _release-gate/ exception.

[github-actions]

⏱️ pre-commit per-hook timing

Hook	Description	Result	Duration
🌐 typecheck — frontend	Whole-project tsc type check	✅	40.01s
🌐 eslint — frontend	Lint + autofix JS/TS	✅	16.38s
📘 typecheck — typescript sdk	Whole-project tsc type check	✅	15.01s
☕ spotless — java backend	Format Java code	✅	7.48s
⚓ helm-docs	Regenerate Helm chart README	✅	3.34s
🐳 hadolint — dockerfiles	Lint Dockerfiles	✅	3.25s
🐍 mypy — python sdk	Static type check	✅	1.60s
⚙️ actionlint — github workflows	Lint GitHub Actions workflows	✅	1.50s
📘 eslint — typescript sdk	Lint + autofix JS/TS	✅	0.96s
🐍 fix end of files — python sdk	Ensure files end in a newline	✅	0.04s
🐍 trim trailing whitespace — python sdk	Strip trailing whitespace	✅	0.04s
🛡️ fix end of files — guardrails	Ensure files end in a newline	✅	0.03s
🐍 ruff-format — python sdk	Format Python code (ruff)	✅	0.02s
🐍 ruff — python sdk	Lint + autofix Python (ruff)	✅	0.02s
🛡️ trim trailing whitespace — guardrails	Strip trailing whitespace	✅	0.02s
block non-public FE plugins	Block non-public FE plugins	✅	0.01s
Total (16 ran)			89.71s

⏭️ 25 skipped (no matching files changed)

Hook	Description	Result
🤖 trim trailing whitespace — optimizer	Strip trailing whitespace	⏭️
🤖 fix end of files — optimizer	Ensure files end in a newline	⏭️
🤖 check yaml — optimizer	Validate YAML syntax	⏭️
🤖 check json — optimizer	Validate JSON syntax	⏭️
🤖 check toml — optimizer	Validate TOML syntax	⏭️
🤖 check for added large files — optimizer	Block large files (>1MB)	⏭️
🔐 detect private key — optimizer	Block committed private keys	⏭️
🤖 check for merge conflicts — optimizer	Block merge-conflict markers	⏭️
🤖 check for case conflicts — optimizer	Block case-only name clashes	⏭️
🤖 pyupgrade — optimizer	Modernize Python syntax	⏭️
🤖 ruff — optimizer	Lint + autofix Python (ruff)	⏭️
🤖 ruff-format — optimizer	Format Python code (ruff)	⏭️
🤖 mypy — optimizer	Static type check	⏭️
📓 nbstripout — optimizer notebooks	Strip notebook output	⏭️
📝 markdownlint — optimizer	Lint Markdown	⏭️
🔤 codespell — optimizer	Fix common misspellings	⏭️
📊 radon cc — optimizer	Cyclomatic-complexity gate	⏭️
📊 radon raw — optimizer	Raw size metrics gate	⏭️
📊 xenon — optimizer	Fail on complexity thresholds	⏭️
📊 lizard — optimizer	Cyclomatic-complexity gate	⏭️
🧹 vulture — optimizer	Find dead code	⏭️
🛡️ ruff — guardrails	Lint + autofix Python (ruff)	⏭️
🛡️ ruff-format — guardrails	Format Python code (ruff)	⏭️
🛡️ mypy — guardrails	Static type check	⏭️
🧪 pre-commit wrapper smoke tests	Self-test the wrapper scripts	⏭️

[AndreiCautisanu]

Reached final version for this PR