agentleFS
Sn4kyGitSn4kyGit/qa-agent-blueprint/.claude/skills/report-to-jira/SKILL.md

report-to-jira

Skill2 starsChanged 8 days agoApache-2.0View on GitHub

Report a test verdict back to JIRA per the active project's PROJECT.md — comment, attachments, workflow transition; Cloud = ADF, Data Center = plain Markdown. Use after a test run completes (PASS, FAIL, BLOCKED, NEEDS_REVIEW), "/report-to-jira", "melde das Ergebnis zurück", "kommentier das Ticket". Not for composing the bug ticket itself — /bug-report.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    No comment, no transition, no attachment — see .claude/rules/core/security.md → "Shadow mode".

    line 17
  2. Very specific:

    Read projects/<KEY>/PROJECT.md → "Reporting": report format (jira-cloud/plain-text), transition names for PASS/FAIL/BLOCKED, and whether …

    line 42
  3. Distinctive:

    When a product was routed, name it in the first line — a verdict without the product is ambiguous once a project has several.

    line 56
  4. Very specific:

    If jira-cloud: build the comment as ADF on a Cloud deployment, or as simple Markdown on a Data Center deployment (the community server …

    line 74
  5. Very specific:

    Put a local path (evidence/..., absolute machine path, file://) into the comment or description — filenames only, the paths belong in the …

    line 133

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/report-to-jira — comment + attachments + transition

Before anything: is this a shadow run?

Shadow mode is on by default. Only when QA_AGENT_SHADOW=0 is set or PROJECT.md → "Reporting" explicitly says Shadow mode: off may you post. Otherwise compose everything exactly as below but post nothing: write the finished comment and the transition you would have applied to evidence/<KEY>/<TICKET>/<ts>/proposed-report.md, set "reported_to_jira": false and "shadow": true in the summary, and tell the user it is waiting for review. No comment, no transition, no attachment — see .claude/rules/core/security.md → "Shadow mode". While shadow is on, the pre-tool guard mechanically blocks the atlassian write tools (including the Cloud v2 meta-tools executeWrite / executeDestructive) — a blocked call is the guard working as designed, not a JIRA error and never a reason to retry or route around it.

NO_TEST posts nothing

A NO_TEST summary means no test completed — there is no verdict to report. Do not comment, do not transition, do not attach. The summary itself is the record (.claude/rules/core/evidence-policy.md).

Before the mechanics: which channel?

PROJECT.md → "JIRA access" owns this decision, read it before anything MCP-shaped below. On browser there is no atlassian MCP server connected at all — every instruction further down that mentions ADF, Markdown conversion, or an MCP attachment call does not apply; the entire post path (comment, attachments, transition) runs through the playwright-driven UI per docs/rules/testing/jira-browser.md → "Writing". A failed Atlassian MCP call call in this mode is not a JIRA outage and not a reason to BLOCK — it is the wrong channel for this project.

Steps

  1. Read projects/<KEY>/PROJECT.md → "Reporting": report format (jira-cloud/plain-text), transition names for PASS/FAIL/BLOCKED, and whether BLOCKED gets a transition at all. Transitions and report format are read from PROJECT.md only, never from a PRODUCT.md — they are properties of the JIRA project's workflow, which every product in that project shares by definition (.claude/rules/core/product-layer.md). A product may override the bug template and the severity mapping, nothing else here.
  2. Compose the comment:
    • The header line carries the run's coordinates: verdict, environment/build, and Mode: <mcp-only|hybrid|cli-first> from the run's execution_summary.json → mode (mandatory since 0.12.0 — whoever reads the report later must know whether the result was observed live or only asserted; an escalated run names its mode and the escalation, per docs/pipeline/06-reporting.md).
    • When a product was routed, name it in the first line — a verdict without the product is ambiguous once a project has several. A ticket covering two products gets one aggregated comment naming both (PASS (STOREFRONT) / BLOCKED (ADMIN), never a bare PASS) and at most one transition, chosen by the worst verdict: FAIL > BLOCKED > NEEDS_REVIEW > PASS.
    • PASS: one-line verdict, bullet list of what was checked, final confirming screenshot attached, evidence files referenced by filename (attached to the ticket).
    • FAIL: reference the bug ticket created via bug-report rather than re-describing the finding inline; keep this comment short.
    • BLOCKED: explain exactly why (env down, missing test data, no deploy permission) and what's needed to unblock — this is often the most useful comment for a human to act on quickly.
    • NEEDS_REVIEW: the test failed but failure-analysis classified the test itself as stale or brittle, not the product. Post the proposal from evidence/<KEY>/<TICKET>/<ts>/failure-analysis.md — including its "do not apply" line and its "if you rule the other way" branch — and say plainly that a person has to decide. Never post a proposal without both branches.
  3. If jira-cloud: build the comment as ADF on a Cloud deployment, or as simple Markdown on a Data Center deployment (the community server converts to wiki markup) — deployment comes from PROJECT.md → "JIRA deployment"; headings/lists, no pipe tables — docs/rules/testing/jira-confluence.md — via the Atlassian MCP server. Attach evidence files (screenshots, not full logs unless the project asks for them) via the attachment call. On Cloud v2 the comment/transition tools are Primary (direct); attachment upload is a DEFERRED tool — discover once, then executeWrite with uploadAttachmentToJiraIssue (docs/rules/testing/jira-confluence.md → "Cloud v2"). On a project with JIRA access: browser, skip ADF/Markup entirely and post through the playwright-driven UI instead — comment, attachments and transition per docs/rules/testing/jira-browser.md → "Writing".
  4. Apply the transition from step 1 — exact name, never guessed. If the named transition doesn't exist in the ticket's current workflow state, stop and report that mismatch rather than trying nearby transition names.

    Which transition depends on the failure class, when failure_analysis is present in the summary:

    ClassTransition
    product-defect, undeterminedthe project's FAIL transition, as before — a possible defect must not pass
    test-outdated, test-brittle, flakecomment only, no transition
    environmentthe project's BLOCKED handling

    The reason the middle row exists: applying Reopen because our assertion held the pre-change contract tears open a developer's correctly-closed ticket, and a QA agent that does that twice is one nobody reads again. The transition names remain sealed to PROJECT.md — this decides whether to transition, never which name to use.

  5. This write goes through the JIRA-write audit automatically — no action needed from you beyond making the actual MCP call. Afterwards update the run's summary (reported_to_jira, shadow) and make sure its skills_used lists the skills the run actually used, report-to-jira included (.claude/rules/core/evidence-policy.md).

Two output channels, two reference rules

Applies to every block this skill composes (comment, bug description, proposal): docs/pipeline/06-reporting.md → "Two output channels". In short:

  • Into the ticket (anything posted or handed to the user for copy-paste): never a local path — no evidence/... run dir, no absolute machine path, no file://. Evidence = filenames + "attached to this ticket". Repo-relative spec paths stay allowed.
  • Into the chat summary to the user: evidence files with their absolute path, one per line — absolute paths are clickable in the harness UI, relative ones lead nowhere half the time. The JIRA comment itself never carries these paths. Afterwards, on interactive macOS sessions, open the run dir in Finder (open "<absolute run-dir>") so the evidence is one preview away — rule and guards (skip on Linux/CI/headless): docs/pipeline/06-reporting.md → "Finder auto-open".

What you never do

Post real credentials, internal URLs, or PII in a JIRA comment. Transition a ticket to a state not defined for this project without checking first. Attach raw evidence/ directory contents wholesale — attach the specific relevant files. Put a local path (evidence/..., absolute machine path, file://) into the comment or description — filenames only, the paths belong in the chat summary.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.