Current job details and evidence
diff --git a/dashboard/public/styles.css b/dashboard/public/styles.css
index c2f36b4..70acf43 100644
--- a/dashboard/public/styles.css
+++ b/dashboard/public/styles.css
@@ -1367,6 +1367,23 @@ pre {
animation: none;
}
}
+#job-summary-dialog {
+ width: min(850px, calc(100vw - 2rem));
+ max-height: calc(100dvh - 2rem);
+ box-sizing: border-box;
+ overflow: auto;
+ overflow-wrap: anywhere;
+ padding: 1.5rem;
+ border: 1px solid #d8d6e8;
+ border-radius: 12px;
+ color: #282638;
+ background: #f7f7fc;
+}
+
+#job-summary-content li {
+ margin-bottom: 0.5rem;
+}
+
#current-job-dialog {
width: 100vw;
max-width: 100vw;
@@ -1380,7 +1397,8 @@ pre {
background: #f7f7fc;
}
-#current-job-dialog::backdrop {
+#current-job-dialog::backdrop,
+#job-summary-dialog::backdrop {
background: rgb(20 18 40 / 45%);
}
diff --git a/dashboard/repository-setup.js b/dashboard/repository-setup.js
index 6280744..afe2f0d 100644
--- a/dashboard/repository-setup.js
+++ b/dashboard/repository-setup.js
@@ -3,7 +3,7 @@ import { dirname, isAbsolute, join, parse, relative, resolve, sep } from "node:p
import { canonicalExecutionRepository } from "./executor-lease.js";
export const SETUP_PROFILE_PATH = join(".github", "fixlab", "repository-profile.json");
-export const REPOSITORY_PROMPT_NAMES = ["bugfix", "intake", "diagnose", "reproduce", "fix", "validate", "live-test", "pr"]
+export const REPOSITORY_PROMPT_NAMES = ["bugfix", "intake", "diagnose", "reproduce", "fix", "validate", "live-test", "pr", "summary"]
.map((name) => `fixlab.${name}.prompt.md`);
export function localRepository(value) {
diff --git a/docs/commands.md b/docs/commands.md
index 6708690..1178e69 100644
--- a/docs/commands.md
+++ b/docs/commands.md
@@ -63,11 +63,35 @@ overwriting repository-owned versions.
| `/fixlab.validate` | Run focused tests, lint, type-check, and builds | Every required check passes or is explicitly skipped with risk |
| `/fixlab.live-test` | Start owned applications and run the browser journey | The running application proves the expected behavior |
| `/fixlab.pr` | Prepare the evidence-backed pull-request outcome | The PR contains the exact change, evidence, and remaining risks |
+| `/fixlab.summary` | Optionally summarize an existing job in the conversation | Actual outcomes, regression gaps, findings, workarounds and must-know items are explicit; no execution or file changes |
The commands read `.github/fixlab/repository-profile.json` and repository
instructions before acting. They never replace repository-owned commands,
environment controls, or approval requirements.
+For an optional engineering handoff, run `/fixlab.summary` in the VS Code
+conversation containing the job evidence. It is never added to the automatic
+bug-fix lifecycle. The prompt returns Markdown without rerunning checks,
+changing job status, creating files or publishing changes. Active-job summaries
+are interim; unavailable results are marked "Not verified". It distinguishes
+runner-confirmed results from agent claims, local from deployed validation, and
+known findings from suspected side effects. A folder or ZIP is a separate request.
+
+Existing repositories can rerun `fixlab init` after updating FixLab to add the
+missing prompt without replacing their profile, agents or customized prompts.
+Other conversation hosts can paste the packaged
+[`fixlab.summary.prompt.md`](../prompts/fixlab.summary.prompt.md) into the
+same evidence-bearing conversation; this is a prompt file, not a new CLI
+subcommand or a dashboard export button.
+
+In the dashboard, select a current or saved job and click **View summary** in
+the header. This separate read-only view assembles recorded investigation,
+changes, gate results, scenario claims, findings and remaining gaps without an
+AI call or job mutation. It refreshes that same job's evidence while open.
+Missing workarounds, commit inventory and side-effect findings stay unverified;
+it does not infer them from a passed status or parse raw logs. For an AI-written
+handoff with richer context, use the optional prompt in its original conversation.
+
## Executor-backed dashboard checks
`fixlab gate --port
--job --gate --repository `
diff --git a/e2e/dashboard-readiness.spec.js b/e2e/dashboard-readiness.spec.js
index fbd53d3..8b58621 100644
--- a/e2e/dashboard-readiness.spec.js
+++ b/e2e/dashboard-readiness.spec.js
@@ -2347,6 +2347,75 @@ function pendingQuestionJob(question) {
return job;
}
+test("job summary shows recorded findings and gaps without execution, and refreshes the same job", async ({ page }) => {
+ const posts = [];
+ page.on("request", (request) => { if (request.method() === "POST") { posts.push(request.url()); } });
+ const job = {
+ id: "summary-current", request: "Investigate a synthetic regression", requestType: "bug-fix",
+ mode: "fix-and-validate", status: "running", targetEnvironment: "dev",
+ stages: {
+ diagnosis: { status: "passed", message: "Confirmed route mismatch" },
+ fix: { status: "pending", message: "Correction not applied" },
+ "live-test": { status: "blocked", message: "Authentication unavailable" }
+ },
+ validation: { scope: "all", gates: [{ id: "regression-tests", status: "failed", required: true, durationMs: 25 }] },
+ testEvidence: { reported: true, scenario: "Check role visibility", source: "synthetic-intercepted", mutationMode: "intercepted" },
+ bugs: [{ id: "bug-1", outcome: "unresolved", owner: "application", summary: "Role selection remains incorrect" }],
+ logs: [{ stream: "stdout", message: "Raw log sentinel must not enter summary" }], activity: []
+ };
+ await mockChatWorkspace(page, { job });
+ await page.goto(baseUrl);
+ await page.getByRole("button", { name: "View summary", exact: true }).click();
+ const summary = page.locator("#job-summary-content");
+ await expect(summary).toContainText("Interim outcome");
+ await expect(summary).toContainText("Confirmed route mismatch");
+ await expect(summary).toContainText("regression-tests: failed (required)");
+ await expect(summary).toContainText("Role selection remains incorrect");
+ await expect(summary).toContainText("Synthetic or intercepted tests do not prove real backend writes");
+ await expect(summary).toContainText("Validated workaround, tradeoffs and fix-induced side effects: Not verified");
+ await expect(summary).not.toContainText("Raw log sentinel");
+ job.status = "passed";
+ job.validation.gates[0].status = "stale";
+ await expect(summary).toContainText("Recorded outcome");
+ await expect(summary).toContainText("regression-tests: stale (required)");
+ await expect(summary).toContainText("Job completion is not a claim of complete regression coverage");
+ await page.getByRole("button", { name: "Close summary" }).click();
+ await expect(page.locator("#job-summary-dialog")).toBeHidden();
+ await expect(page.getByRole("button", { name: "View summary", exact: true })).toBeFocused();
+ expect(posts).toEqual([]);
+});
+
+test("job summary follows historical selection, handles missing evidence and renders text safely", async ({ page }) => {
+ const posts = [];
+ page.on("request", (request) => { if (request.method() === "POST") { posts.push(request.url()); } });
+ const job = { id: "summary-active", request: "Active summary sentinel", requestType: "bug-fix",
+ mode: "fix-and-validate", status: "running", stages: {}, bugs: [], logs: [], activity: [] };
+ const saved = { ...job, id: "summary-history", status: "held",
+ request: 'Saved 
https://private.example.test/path',
+ stages: { review: { status: "blocked", message: "authorization=private-sentinel" } } };
+ await mockChatWorkspace(page, { job, history: [saved] });
+ await page.goto(baseUrl);
+ await page.locator('.workspace-navigation [data-open="memory-group"]').click();
+ await page.locator("#repository-memory-list button").click();
+ await page.setViewportSize({ width: 390, height: 844 });
+ await page.getByRole("button", { name: "View summary", exact: true }).click();
+ const summary = page.locator("#job-summary-content");
+ await expect(summary).toContainText(saved.id);
+ await expect(summary).not.toContainText(job.request);
+ await expect(summary).toContainText("No executor-backed gate results recorded");
+ await expect(summary).toContainText("[Sensitive detail omitted]");
+ await expect(summary).not.toContainText("private.example.test");
+ await expect(summary).not.toContainText("private-sentinel");
+ await expect(summary.locator("img")).toHaveCount(0);
+ expect(await page.evaluate(() => window.summaryInjection)).toBeUndefined();
+ const overflow = await page.locator("#job-summary-dialog").evaluate((element) =>
+ element.scrollWidth > element.clientWidth);
+ expect(overflow).toBe(false);
+ await page.keyboard.press("Escape");
+ await expect(page.locator("#job-summary-dialog")).toBeHidden();
+ expect(posts).toEqual([]);
+});
+
test("current-job full screen ignores historical selection, follows the live job and restores the prior view", async ({ page }) => {
const posts = [];
let historicalEvidence;
diff --git a/prompts/fixlab.summary.prompt.md b/prompts/fixlab.summary.prompt.md
new file mode 100644
index 0000000..15b25f5
--- /dev/null
+++ b/prompts/fixlab.summary.prompt.md
@@ -0,0 +1,47 @@
+---
+mode: agent
+description: Optionally summarize an existing FixLab job without running checks or changing files.
+---
+
+Generate an engineering handoff for the FixLab job in this conversation using
+existing session evidence only. This is an optional summary, not a new fix or
+validation job. If the job is ambiguous or its evidence is unavailable, ask for
+the job identifier and relevant sanitized results; do not infer them from a
+different job.
+
+Do not rerun tests, install tools, start applications, access live systems,
+change files, commit, publish, or alter job status. Do not invoke the bug-fix
+workflow. Return Markdown in the conversation only; a folder or ZIP requires
+a separate explicit request and normal permissions.
+
+Include these sections, marking missing evidence "Not verified":
+
+1. Outcome: original bug, expected versus actual behavior, job identifier,
+ repository/commit and PR references when safe, and whether the result is
+ fixed and verified, fixed but unverified, workaround-only, external, or
+ unresolved. For an active job, label the summary interim.
+2. Diagnosis and changes: evidence-backed root cause, affected components,
+ changed files and the rationale for the correction. Explain no-change
+ outcomes honestly.
+3. Test and regression report: a table of scenarios/assertions, exact safe
+ commands and working directories, recorded outcomes/durations, and evidence
+ references. Separate the original failing reproduction from subsequent
+ verification, focused tests from broad regression, and runner-confirmed
+ results from agent-reported claims. Preserve failed, skipped, timed-out,
+ blocked, pending, stale, not-run and deferred checks; never mark them passed.
+4. Findings and must-know: verified tradeoffs, compatibility concerns,
+ maintenance traps and prerequisites. Separate pre-existing defects from
+ confirmed fix-induced regressions; label suspected side effects as
+ hypotheses, not findings.
+5. Workarounds: only recorded workarounds, their validation, limitations,
+ conditions for removal, and known ownership. Do not invent one.
+6. Remaining risks and next actions: untested scenarios, missing data/access,
+ evidence gaps and concrete follow-up. A complete summary does not imply
+ complete regression coverage.
+
+Distinguish local changed-code validation from deployed-system checks,
+synthetic/intercepted tests from real integration, and draft publication from
+ready-for-review or merge. Reference existing reviewed evidence instead of
+pasting raw logs or transcripts. Exclude credentials, authorization headers,
+cookies, browser state, private endpoints, personal/customer data, unreviewed
+screenshots and model-cost/token metrics. Omit or redact sensitive references.
diff --git a/test/dashboard-repository-settings.test.js b/test/dashboard-repository-settings.test.js
index 4ab2c83..26c636f 100644
--- a/test/dashboard-repository-settings.test.js
+++ b/test/dashboard-repository-settings.test.js
@@ -32,7 +32,7 @@ test("repository onboarding previews without writes and creates only missing rev
assert.equal(localRepository(repository), repository);
const preview = plan(repository);
assert.equal(existsSync(join(repository, ".github")), false);
- assert.equal(preview.files.length, 10);
+ assert.equal(preview.files.length, 11);
const applied = applyRepositorySettings(preview, "synthetic-settings-owner");
assert.equal(applied.saved, true);
const profile = JSON.parse(readFileSync(join(repository, SETUP_PROFILE_PATH), "utf8"));
@@ -46,6 +46,10 @@ test("repository onboarding previews without writes and creates only missing rev
assert.equal(configuredTestSynthesis(profile, "", undefined, "sit").source, "synthetic-intercepted");
assert.equal(validateLiveTestProfile(profile).ok, true);
assert.equal(existsSync(join(repository, ".github", "agents", "fixlab-autofix.agent.md")), true);
+ assert.equal(
+ readFileSync(join(repository, ".github", "prompts", "fixlab.summary.prompt.md"), "utf8"),
+ readFileSync(join(packageRoot, "prompts", "fixlab.summary.prompt.md"), "utf8")
+ );
assert.equal(existsSync(canonicalExecutionRepository(repository).directory), false);
assert.equal(plan(repository).assets.length, 0);
} finally {
diff --git a/test/fixlab-cli.test.js b/test/fixlab-cli.test.js
index d6b2d07..5425516 100644
--- a/test/fixlab-cli.test.js
+++ b/test/fixlab-cli.test.js
@@ -422,7 +422,8 @@ test("init creates a parseable repository profile", () => {
"fixlab.fix.prompt.md",
"fixlab.validate.prompt.md",
"fixlab.live-test.prompt.md",
- "fixlab.pr.prompt.md"
+ "fixlab.pr.prompt.md",
+ "fixlab.summary.prompt.md"
];
for (const promptName of expectedPrompts) {
assert.equal(existsSync(join(promptDirectory, promptName)), true);
@@ -446,6 +447,10 @@ test("init creates a parseable repository profile", () => {
);
assert.match(bugfixPrompt, /agent: fixlab-autofix/);
assert.match(bugfixPrompt, /smallest complete correction/);
+ assert.equal(
+ readFileSync(join(promptDirectory, "fixlab.summary.prompt.md"), "utf8"),
+ readFileSync(new URL("../prompts/fixlab.summary.prompt.md", import.meta.url), "utf8")
+ );
} finally {
rmSync(repository, { recursive: true, force: true });
}
@@ -466,7 +471,7 @@ test("init preserves existing profile and prompt files", () => {
repository,
".github",
"prompts",
- "fixlab.validate.prompt.md"
+ "fixlab.summary.prompt.md"
);
const agentPath = join(
repository,