lastlight 0.1.15 → 0.2.1

package/workflows/security-feedback.yaml

@@ -0,0 +1,25 @@
+kind: health
+name: security-feedback
+description: |
+  Process maintainer feedback on a security-labelled issue (usually a
+  per-run scan summary). Classifies the comment and acts:
+
+    - create-issues — break selected findings out into individual issues
+      that can later be fed to the `build` skill (the primary flow)
+    - accept-risk / false-positive — update SECURITY.md via PR to
+      suppress the finding in future scans
+    - reopen — ask the maintainer to re-run `@last-light security-review`
+    - discuss / ignore — reply conversationally or do nothing
+
+  The scan-summary issue grammar is defined in
+  `skills/security-review/SKILL.md § Issue format`; this skill's parser
+  mirrors it (version-checked on entry).
+
+  Triggered by:
+    - @last-light mention on issues labelled "security"
+
+phases:
+  - name: feedback
+    label: Security feedback
+    skill: security-feedback
+    model: "{{models.security}}"

package/workflows/security-review.yaml

@@ -0,0 +1,29 @@
+kind: health
+name: security-review
+description: |
+  Scan the target repository for security issues using open-source tools
+  (npm audit, semgrep, gitleaks) plus AI code review. Files ONE summary
+  issue per run (Renovate-style) containing a GitHub task list of every
+  finding grouped by Critical / High / Medium / Low. Honours SECURITY.md
+  for known false positives and accepted risks.
+
+  Maintainers break findings out into individual issues by commenting on
+  the summary issue (see the `security-feedback` skill — which handles
+  `@last-light create issues for the criticals`, per-item accept-risk /
+  false-positive commands, and conversational follow-ups).
+
+  Each invocation scans exactly ONE repo. The weekly cron fans out into
+  one independent run per managed repo.
+
+  Triggered by:
+    - weekly cron (Monday 10:00, fans out over every managed repo)
+    - manual CLI: tsx src/cli.ts security owner/repo
+    - @last-light security-review in a GitHub comment
+    - Slack: "security review cliftonc/repo"
+
+phases:
+  - name: scan
+    label: Security scan
+    skill: security-review
+    model: "{{models.security}}"
+    variant: "{{variants.security}}"

package/workflows/verify.yaml

@@ -0,0 +1,85 @@
+kind: verify
+name: verify
+description: |
+  Verify a behaviour claim against a repo or PR and report whether the evidence
+  confirms or refutes it. It installs and runs the code in the sandbox, captures
+  bash/text evidence, optionally drives the UI in a real browser, then posts a
+  single CONFIRMED / REFUTED / INCONCLUSIVE verdict. Does not modify code.
+
+  Delivery is a single in-place status comment (`status_checklist`): the run
+  shows its phases ticking off, then a final `synthesize` phase folds the text
+  verdict and (when it ran) the browser evidence into ONE combined verdict,
+  delivered as the comment's footer via `final_message`. No comment-per-phase.
+  Posted on the originating issue/PR for GitHub runs, or into the Slack thread
+  for Slack runs. The agents therefore do NOT post anything themselves.
+
+  When the host runs the docker backend with the browser-QA image built, the
+  gated `verify_browser` phase drives the UI in a real headless Chromium and
+  captures screenshot evidence; everywhere else it is silently skipped (graceful
+  degradation) and the text verdict stands alone. The synthesis runs either way.
+
+  Triggered by:
+    - a `@last-light verify <claim>` comment on an issue/PR
+    - manual CLI: tsx src/cli.ts verify owner/repo#N -- "<claim>"
+
+status_checklist: true
+final_message: "{{synthesisResult}}"
+
+phases:
+  - name: phase_0
+    label: Context
+    type: context
+    depends_on: []
+
+  - name: verify
+    label: Verify
+    depends_on: [phase_0]
+    prompt: prompts/verify.md
+    # verify: the investigator procedure + report shape (primary).
+    # building: install + run the repo to capture evidence.
+    skills: [verify, building]
+    model: "{{models.verify}}"
+    variant: "{{variants.verify}}"
+    output_var: verifyResult
+    messages:
+      on_start: "Verifying the claim — installing, running, and capturing evidence..."
+      on_success: "Text verification complete."
+      on_failure: "I couldn't complete the verification — see the logs for details."
+
+  # Browser-QA pass (Tier B). Gated to the docker backend AND the enriched
+  # `lastlight-sandbox-qa:latest` image (Playwright + Chromium). On gondolin, or
+  # a docker host without that image, the scheduler silently skips this phase.
+  # It is terminal (nothing depends on it) — `synthesize` depends only on
+  # `verify`, so it still runs (after this phase, by declaration order) and
+  # folds in `verifyBrowserResult` only when this phase actually produced it.
+  - name: verify_browser
+    label: Verify (browser)
+    depends_on: [verify]
+    prompt: prompts/verify-browser.md
+    skills: [verify, building, browser-qa]
+    model: "{{models.verify}}"
+    variant: "{{variants.verify}}"
+    requires_sandbox: docker
+    sandbox_image: qa
+    output_var: verifyBrowserResult
+    messages:
+      on_start: "Browser QA available — driving the UI and capturing screenshots..."
+      on_success: "Browser QA complete — screenshots captured."
+      on_skipped_done: "Browser QA unavailable on this host — text verdict stands."
+
+  # Synthesis. Depends only on `verify` (always runs after the gated browser
+  # phase by declaration order, and never skips when that phase is gated out).
+  # Reads the prior reports from phaseOutputs and writes ONE concise combined
+  # verdict; its output (`synthesisResult`) becomes the comment footer via the
+  # workflow-level `final_message`. Pure composition — no skills, no repo work.
+  - name: synthesize
+    label: Summary
+    depends_on: [verify]
+    prompt: prompts/verify-synth.md
+    model: "{{models.verify}}"
+    variant: "{{variants.verify}}"
+    output_var: synthesisResult
+    messages:
+      on_start: "Synthesizing the final verdict..."
+      on_success: "Verdict posted below."
+      on_failure: "I couldn't synthesize the final verdict — see the logs."

package/deploy/mcp-config.tmpl.json

@@ -1,14 +0,0 @@
-{
-  "mcpServers": {
-    "github": {
-      "command": "node",
-      "args": ["/app/mcp-github-app/src/index.js"],
-      "env": {
-        "GITHUB_APP_ID": "${GITHUB_APP_ID}",
-        "GITHUB_APP_PRIVATE_KEY_PATH": "${GITHUB_APP_PRIVATE_KEY_PATH}",
-        "GITHUB_APP_INSTALLATION_ID": "${GITHUB_APP_INSTALLATION_ID}",
-        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
-      }
-    }
-  }
-}

package/dist/cron/rate-limits.d.ts

@@ -1,26 +0,0 @@
-import type { StateDb } from "../state/db.js";
-/** Component name used for the host claude CLI auth state in system_status. */
-export declare const HOST_CLAUDE_AUTH_COMPONENT = "host-claude-auth";
-/**
- * Optional notifier — called once when the auth state transitions to degraded.
- * Pass a SlackConnector's `sendToDeliveryChannel` (or any other channel).
- */
-export type AdminNotifier = (text: string) => Promise<void>;
-/**
- * Check Claude usage limits by making a minimal Agent SDK call
- * and parsing the stream-json output for rate_limit_event and usage data.
- *
- * Works with Claude subscriptions (no API key needed).
- *
- * Behavior:
- * - If the host claude CLI returns "Not logged in" / authentication_failed,
- *   marks `system_status.host-claude-auth` as `degraded`, notifies the admin
- *   on the first transition, and halts subsequent runs until cleared.
- * - The capacity check creates a JSONL session file as a side effect — we
- *   delete it after parsing so it doesn't pollute the dashboard Sessions list.
- */
-export interface CheckApiUsageOptions {
-    /** Bypass the auth-degraded early return — used by manual rechecks. */
-    force?: boolean;
-}
-export declare function checkApiUsage(db: StateDb, notifyAdmin?: AdminNotifier, opts?: CheckApiUsageOptions): Promise<void>;

package/dist/cron/rate-limits.js

@@ -1,193 +0,0 @@
-/** Component name used for the host claude CLI auth state in system_status. */
-export const HOST_CLAUDE_AUTH_COMPONENT = "host-claude-auth";
-export async function checkApiUsage(db, notifyAdmin, opts = {}) {
-    // Halt early if we're already in a degraded state — operator must clear it
-    // (e.g. after `claude /login` + recheck) before this cron resumes.
-    // The recheck endpoint passes force:true to bypass this.
-    const existing = db.getSystemStatus(HOST_CLAUDE_AUTH_COMPONENT);
-    if (!opts.force && existing?.state === "degraded") {
-        console.log(`[usage] Skipping capacity check — host claude auth is degraded since ${existing.since}. Run claude /login on the host and trigger a recheck to resume.`);
-        return;
-    }
-    const now = new Date().toISOString();
-    let capacityCheckSessionId = "";
-    let authFailed = false;
-    let authFailureReason = "";
-    // ── Capacity & rate limit check via claude CLI ──
-    try {
-        const { execFile } = await import("child_process");
-        const { promisify } = await import("util");
-        const exec = promisify(execFile);
-        let stdout = "";
-        try {
-            // NOTE: do NOT pass --bare here. --bare disables OAuth credential
-            // reading and requires ANTHROPIC_API_KEY, which we don't have on this
-            // host (we authenticate via `claude /login` to a Claude subscription).
-            const result = await exec("claude", [
-                "--print", "reply with just: ok",
-                "--output-format", "stream-json",
-                "--verbose",
-                "--max-turns", "1",
-                "--model", process.env.CLAUDE_MODEL || "claude-sonnet-4-6",
-            ], {
-                timeout: 30_000,
-                env: { ...process.env, HOME: process.env.HOME || "/home/lastlight" },
-            });
-            stdout = result.stdout;
-        }
-        catch (execErr) {
-            // Non-zero exit (e.g. auth_failed) — claude still emits stream-json on stdout
-            stdout = execErr.stdout?.toString() || "";
-            if (!stdout) {
-                // No output at all — propagate the underlying failure
-                throw execErr;
-            }
-        }
-        // Parse each JSON line from the stream output
-        for (const line of stdout.split("\n")) {
-            if (!line.startsWith("{"))
-                continue;
-            try {
-                const msg = JSON.parse(line);
-                // Capture session id so we can delete the JSONL file afterwards
-                if (msg.type === "system" && msg.subtype === "init" && msg.session_id) {
-                    capacityCheckSessionId = msg.session_id;
-                }
-                // Detect "Not logged in" — claude emits a synthetic assistant message
-                // with error: "authentication_failed" when the host CLI has no creds.
-                if (msg.type === "assistant" && msg.error === "authentication_failed") {
-                    authFailed = true;
-                    const text = msg.message?.content?.[0]?.text || "Not logged in";
-                    authFailureReason = text;
-                }
-                // Rate limit info from subscription
-                if (msg.type === "rate_limit_event" && msg.rate_limit_info) {
-                    const info = msg.rate_limit_info;
-                    db.updateRateLimit("subscription:status", info.status === "allowed" ? 1 : 0, now);
-                    db.updateRateLimit("subscription:type", 0, info.rateLimitType || "unknown");
-                    if (info.resetsAt) {
-                        db.updateRateLimit("subscription:resets_at", info.resetsAt, new Date(info.resetsAt * 1000).toISOString());
-                    }
-                    if (info.overageStatus) {
-                        db.updateRateLimit("subscription:overage_status", info.overageStatus === "allowed" ? 1 : 0, now);
-                    }
-                    if (info.overageResetsAt) {
-                        db.updateRateLimit("subscription:overage_resets_at", info.overageResetsAt, new Date(info.overageResetsAt * 1000).toISOString());
-                    }
-                    db.updateRateLimit("subscription:using_overage", info.isUsingOverage ? 1 : 0, now);
-                    console.log(`[usage] Subscription: ${info.status}, type=${info.rateLimitType}, overage=${info.isUsingOverage ? "yes" : "no"} (${info.overageStatus})`);
-                }
-                // Result with cost and usage breakdown
-                if (msg.type === "result") {
-                    if (msg.total_cost_usd !== undefined) {
-                        // Store as cents to keep as integer
-                        db.updateRateLimit("usage:last_check_cost_cents", Math.round(msg.total_cost_usd * 100), now);
-                    }
-                    if (msg.usage) {
-                        db.updateRateLimit("usage:last_check_input_tokens", msg.usage.input_tokens || 0, now);
-                        db.updateRateLimit("usage:last_check_output_tokens", msg.usage.output_tokens || 0, now);
-                    }
-                }
-            }
-            catch { /* skip unparseable lines */ }
-        }
-    }
-    catch (err) {
-        // The capacity check is best-effort — log and continue. Cleanup + status
-        // handling below should still run regardless of the failure mode.
-        db.updateRateLimit("subscription:status", 0, now);
-        console.warn(`[usage] Capacity check failed: ${err.message}`);
-    }
-    // If the call surfaced an auth failure, mark the component degraded and
-    // notify the admin once on transition. Subsequent runs early-return above.
-    if (authFailed) {
-        const transitioned = db.setSystemStatus(HOST_CLAUDE_AUTH_COMPONENT, "degraded", authFailureReason || "Host claude CLI is not logged in");
-        if (transitioned) {
-            console.error(`[usage] Host claude auth is now DEGRADED: ${authFailureReason}`);
-            if (notifyAdmin) {
-                const msg = `:rotating_light: *Last Light: host claude CLI auth degraded*\n` +
-                    `Reason: ${authFailureReason}\n` +
-                    `Action: run \`docker exec -it --user lastlight lastlight-agent-1 claude /login\` and then trigger a recheck.\n` +
-                    `The capacity-check cron is now halted to prevent log spam.`;
-                try {
-                    await notifyAdmin(msg);
-                }
-                catch (notifyErr) {
-                    console.error(`[usage] Failed to notify admin: ${notifyErr.message}`);
-                }
-            }
-        }
-    }
-    else {
-        // Successful run — clear any prior degraded state
-        const existing = db.getSystemStatus(HOST_CLAUDE_AUTH_COMPONENT);
-        if (existing && existing.state !== "ok") {
-            db.setSystemStatus(HOST_CLAUDE_AUTH_COMPONENT, "ok");
-            console.log(`[usage] Host claude auth recovered — state cleared`);
-            if (notifyAdmin) {
-                try {
-                    await notifyAdmin(`:white_check_mark: Last Light: host claude CLI auth recovered. Capacity-check cron resuming normal operation.`);
-                }
-                catch { /* non-fatal */ }
-            }
-        }
-        else if (!existing) {
-            // First-ever successful run — initialize the state
-            db.setSystemStatus(HOST_CLAUDE_AUTH_COMPONENT, "ok");
-        }
-    }
-    // Run JSONL cleanup regardless of outcome (try/finally semantics).
-    {
-        // Delete the session JSONL file the capacity check created so it doesn't
-        // pollute the dashboard Sessions list. Best-effort — ignore failures.
-        if (capacityCheckSessionId) {
-            try {
-                const { unlink } = await import("fs/promises");
-                const { join } = await import("path");
-                const home = process.env.HOME || "/home/lastlight";
-                // claude encodes the project's cwd as the directory name, replacing
-                // path separators with dashes. e.g. /app -> -app
-                const projectDir = process.cwd().replace(/\//g, "-");
-                const sessionFile = join(home, ".claude", "projects", projectDir, `${capacityCheckSessionId}.jsonl`);
-                await unlink(sessionFile);
-            }
-            catch {
-                /* file may not exist or claude may not have written it — ignore */
-            }
-        }
-    }
-    // ── Usage stats from our own execution records ──
-    try {
-        const stats = computeUsageStats(db);
-        db.updateRateLimit("usage:executions_1h", stats.lastHour, now);
-        db.updateRateLimit("usage:executions_24h", stats.last24h, now);
-        db.updateRateLimit("usage:turns_1h", stats.turnsLastHour, now);
-        db.updateRateLimit("usage:turns_24h", stats.turnsLast24h, now);
-        console.log(`[usage] Last 1h: ${stats.lastHour} executions, ${stats.turnsLastHour} turns | ` +
-            `24h: ${stats.last24h} executions, ${stats.turnsLast24h} turns`);
-    }
-    catch (err) {
-        console.error(`[usage] Stats computation failed: ${err.message}`);
-    }
-}
-function computeUsageStats(db) {
-    const now = Date.now();
-    const oneHourAgo = new Date(now - 60 * 60 * 1000).toISOString();
-    const oneDayAgo = new Date(now - 24 * 60 * 60 * 1000).toISOString();
-    const raw = db.database;
-    const h1 = raw.prepare(`
-    SELECT COUNT(*) as count, COALESCE(SUM(turns), 0) as turns
-    FROM executions WHERE started_at >= ?
-  `).get(oneHourAgo);
-    const h24 = raw.prepare(`
-    SELECT COUNT(*) as count, COALESCE(SUM(turns), 0) as turns
-    FROM executions WHERE started_at >= ?
-  `).get(oneDayAgo);
-    return {
-        lastHour: h1.count,
-        turnsLastHour: h1.turns,
-        last24h: h24.count,
-        turnsLast24h: h24.turns,
-    };
-}
-//# sourceMappingURL=rate-limits.js.map

package/dist/cron/rate-limits.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"rate-limits.js","sourceRoot":"","sources":["../../src/cron/rate-limits.ts"],"names":[],"mappings":"AAEA,+EAA+E;AAC/E,MAAM,CAAC,MAAM,0BAA0B,GAAG,kBAAkB,CAAC;AA0B7D,MAAM,CAAC,KAAK,UAAU,aAAa,CACjC,EAAW,EACX,WAA2B,EAC3B,OAA6B,EAAE;IAE/B,2EAA2E;IAC3E,mEAAmE;IACnE,yDAAyD;IACzD,MAAM,QAAQ,GAAG,EAAE,CAAC,eAAe,CAAC,0BAA0B,CAAC,CAAC;IAChE,IAAI,CAAC,IAAI,CAAC,KAAK,IAAI,QAAQ,EAAE,KAAK,KAAK,UAAU,EAAE,CAAC;QAClD,OAAO,CAAC,GAAG,CAAC,wEAAwE,QAAQ,CAAC,KAAK,kEAAkE,CAAC,CAAC;QACtK,OAAO;IACT,CAAC;IAED,MAAM,GAAG,GAAG,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IACrC,IAAI,sBAAsB,GAAG,EAAE,CAAC;IAChC,IAAI,UAAU,GAAG,KAAK,CAAC;IACvB,IAAI,iBAAiB,GAAG,EAAE,CAAC;IAE3B,mDAAmD;IACnD,IAAI,CAAC;QACH,MAAM,EAAE,QAAQ,EAAE,GAAG,MAAM,MAAM,CAAC,eAAe,CAAC,CAAC;QACnD,MAAM,EAAE,SAAS,EAAE,GAAG,MAAM,MAAM,CAAC,MAAM,CAAC,CAAC;QAC3C,MAAM,IAAI,GAAG,SAAS,CAAC,QAAQ,CAAC,CAAC;QAEjC,IAAI,MAAM,GAAG,EAAE,CAAC;QAChB,IAAI,CAAC;YACH,kEAAkE;YAClE,sEAAsE;YACtE,uEAAuE;YACvE,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,QAAQ,EAAE;gBAClC,SAAS,EAAE,qBAAqB;gBAChC,iBAAiB,EAAE,aAAa;gBAChC,WAAW;gBACX,aAAa,EAAE,GAAG;gBAClB,SAAS,EAAE,OAAO,CAAC,GAAG,CAAC,YAAY,IAAI,mBAAmB;aAC3D,EAAE;gBACD,OAAO,EAAE,MAAM;gBACf,GAAG,EAAE,EAAE,GAAG,OAAO,CAAC,GAAG,EAAE,IAAI,EAAE,OAAO,CAAC,GAAG,CAAC,IAAI,IAAI,iBAAiB,EAAE;aACrE,CAAC,CAAC;YACH,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QACzB,CAAC;QAAC,OAAO,OAAY,EAAE,CAAC;YACtB,8EAA8E;YAC9E,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC;YAC1C,IAAI,CAAC,MAAM,EAAE,CAAC;gBACZ,sDAAsD;gBACtD,MAAM,OAAO,CAAC;YAChB,CAAC;QACH,CAAC;QAED,8CAA8C;QAC9C,KAAK,MAAM,IAAI,IAAI,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;YACtC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC;gBAAE,SAAS;YACpC,IAAI,CAAC;gBACH,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAE7B,gEAAgE;gBAChE,IAAI,GAAG,CAAC,IAAI,KAAK,QAAQ,IAAI,GAAG,CAAC,OAAO,KAAK,MAAM,IAAI,GAAG,CAAC,UAAU,EAAE,CAAC;oBACtE,sBAAsB,GAAG,GAAG,CAAC,UAAU,CAAC;gBAC1C,CAAC;gBAED,sEAAsE;gBACtE,sEAAsE;gBACtE,IAAI,GAAG,CAAC,IAAI,KAAK,WAAW,IAAI,GAAG,CAAC,KAAK,KAAK,uBAAuB,EAAE,CAAC;oBACtE,UAAU,GAAG,IAAI,CAAC;oBAClB,MAAM,IAAI,GAAG,GAAG,CAAC,OAAO,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,IAAI,eAAe,CAAC;oBAChE,iBAAiB,GAAG,IAAI,CAAC;gBAC3B,CAAC;gBAED,oCAAoC;gBACpC,IAAI,GAAG,CAAC,IAAI,KAAK,kBAAkB,IAAI,GAAG,CAAC,eAAe,EAAE,CAAC;oBAC3D,MAAM,IAAI,GAAG,GAAG,CAAC,eAAe,CAAC;oBACjC,EAAE,CAAC,eAAe,CAAC,qBAAqB,EAAE,IAAI,CAAC,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;oBAClF,EAAE,CAAC,eAAe,CAAC,mBAAmB,EAAE,CAAC,EAAE,IAAI,CAAC,aAAa,IAAI,SAAS,CAAC,CAAC;oBAE5E,IAAI,IAAI,CAAC,QAAQ,EAAE,CAAC;wBAClB,EAAE,CAAC,eAAe,CAAC,wBAAwB,EAAE,IAAI,CAAC,QAAQ,EAAE,IAAI,IAAI,CAAC,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC;oBAC5G,CAAC;oBACD,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;wBACvB,EAAE,CAAC,eAAe,CAAC,6BAA6B,EAAE,IAAI,CAAC,aAAa,KAAK,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;oBACnG,CAAC;oBACD,IAAI,IAAI,CAAC,eAAe,EAAE,CAAC;wBACzB,EAAE,CAAC,eAAe,CAAC,gCAAgC,EAAE,IAAI,CAAC,eAAe,EAAE,IAAI,IAAI,CAAC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC;oBAClI,CAAC;oBACD,EAAE,CAAC,eAAe,CAAC,4BAA4B,EAAE,IAAI,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;oBAEnF,OAAO,CAAC,GAAG,CAAC,yBAAyB,IAAI,CAAC,MAAM,UAAU,IAAI,CAAC,aAAa,aAAa,IAAI,CAAC,cAAc,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,IAAI,CAAC,aAAa,GAAG,CAAC,CAAC;gBACzJ,CAAC;gBAED,uCAAuC;gBACvC,IAAI,GAAG,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;oBAC1B,IAAI,GAAG,CAAC,cAAc,KAAK,SAAS,EAAE,CAAC;wBACrC,oCAAoC;wBACpC,EAAE,CAAC,eAAe,CAAC,6BAA6B,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,cAAc,GAAG,GAAG,CAAC,EAAE,GAAG,CAAC,CAAC;oBAC/F,CAAC;oBACD,IAAI,GAAG,CAAC,KAAK,EAAE,CAAC;wBACd,EAAE,CAAC,eAAe,CAAC,+BAA+B,EAAE,GAAG,CAAC,KAAK,CAAC,YAAY,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC;wBACtF,EAAE,CAAC,eAAe,CAAC,gCAAgC,EAAE,GAAG,CAAC,KAAK,CAAC,aAAa,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC;oBAC1F,CAAC;gBACH,CAAC;YACH,CAAC;YAAC,MAAM,CAAC,CAAC,4BAA4B,CAAC,CAAC;QAC1C,CAAC;IACH,CAAC;IAAC,OAAO,GAAQ,EAAE,CAAC;QAClB,yEAAyE;QACzE,kEAAkE;QAClE,EAAE,CAAC,eAAe,CAAC,qBAAqB,EAAE,CAAC,EAAE,GAAG,CAAC,CAAC;QAClD,OAAO,CAAC,IAAI,CAAC,kCAAkC,GAAG,CAAC,OAAO,EAAE,CAAC,CAAC;IAChE,CAAC;IAED,wEAAwE;IACxE,2EAA2E;IAC3E,IAAI,UAAU,EAAE,CAAC;QACf,MAAM,YAAY,GAAG,EAAE,CAAC,eAAe,CACrC,0BAA0B,EAC1B,UAAU,EACV,iBAAiB,IAAI,kCAAkC,CACxD,CAAC;QACF,IAAI,YAAY,EAAE,CAAC;YACjB,OAAO,CAAC,KAAK,CAAC,6CAA6C,iBAAiB,EAAE,CAAC,CAAC;YAChF,IAAI,WAAW,EAAE,CAAC;gBAChB,MAAM,GAAG,GACP,gEAAgE;oBAChE,WAAW,iBAAiB,IAAI;oBAChC,gHAAgH;oBAChH,4DAA4D,CAAC;gBAC/D,IAAI,CAAC;oBACH,MAAM,WAAW,CAAC,GAAG,CAAC,CAAC;gBACzB,CAAC;gBAAC,OAAO,SAAc,EAAE,CAAC;oBACxB,OAAO,CAAC,KAAK,CAAC,mCAAmC,SAAS,CAAC,OAAO,EAAE,CAAC,CAAC;gBACxE,CAAC;YACH,CAAC;QACH,CAAC;IACH,CAAC;SAAM,CAAC;QACN,kDAAkD;QAClD,MAAM,QAAQ,GAAG,EAAE,CAAC,eAAe,CAAC,0BAA0B,CAAC,CAAC;QAChE,IAAI,QAAQ,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,EAAE,CAAC;YACxC,EAAE,CAAC,eAAe,CAAC,0BAA0B,EAAE,IAAI,CAAC,CAAC;YACrD,OAAO,CAAC,GAAG,CAAC,oDAAoD,CAAC,CAAC;YAClE,IAAI,WAAW,EAAE,CAAC;gBAChB,IAAI,CAAC;oBACH,MAAM,WAAW,CAAC,+GAA+G,CAAC,CAAC;gBACrI,CAAC;gBAAC,MAAM,CAAC,CAAC,eAAe,CAAC,CAAC;YAC7B,CAAC;QACH,CAAC;aAAM,IAAI,CAAC,QAAQ,EAAE,CAAC;YACrB,mDAAmD;YACnD,EAAE,CAAC,eAAe,CAAC,0BAA0B,EAAE,IAAI,CAAC,CAAC;QACvD,CAAC;IACH,CAAC;IAED,mEAAmE;IACnE,CAAC;QACC,yEAAyE;QACzE,sEAAsE;QACtE,IAAI,sBAAsB,EAAE,CAAC;YAC3B,IAAI,CAAC;gBACH,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,MAAM,CAAC,aAAa,CAAC,CAAC;gBAC/C,MAAM,EAAE,IAAI,EAAE,GAAG,MAAM,MAAM,CAAC,MAAM,CAAC,CAAC;gBACtC,MAAM,IAAI,GAAG,OAAO,CAAC,GAAG,CAAC,IAAI,IAAI,iBAAiB,CAAC;gBACnD,oEAAoE;gBACpE,iDAAiD;gBACjD,MAAM,UAAU,GAAG,OAAO,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;gBACrD,MAAM,WAAW,GAAG,IAAI,CAAC,IAAI,EAAE,SAAS,EAAE,UAAU,EAAE,UAAU,EAAE,GAAG,sBAAsB,QAAQ,CAAC,CAAC;gBACrG,MAAM,MAAM,CAAC,WAAW,CAAC,CAAC;YAC5B,CAAC;YAAC,MAAM,CAAC;gBACP,mEAAmE;YACrE,CAAC;QACH,CAAC;IACH,CAAC;IAED,mDAAmD;IACnD,IAAI,CAAC;QACH,MAAM,KAAK,GAAG,iBAAiB,CAAC,EAAE,CAAC,CAAC;QACpC,EAAE,CAAC,eAAe,CAAC,qBAAqB,EAAE,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,CAAC;QAC/D,EAAE,CAAC,eAAe,CAAC,sBAAsB,EAAE,KAAK,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;QAC/D,EAAE,CAAC,eAAe,CAAC,gBAAgB,EAAE,KAAK,CAAC,aAAa,EAAE,GAAG,CAAC,CAAC;QAC/D,EAAE,CAAC,eAAe,CAAC,iBAAiB,EAAE,KAAK,CAAC,YAAY,EAAE,GAAG,CAAC,CAAC;QAC/D,OAAO,CAAC,GAAG,CACT,oBAAoB,KAAK,CAAC,QAAQ,gBAAgB,KAAK,CAAC,aAAa,WAAW;YAChF,QAAQ,KAAK,CAAC,OAAO,gBAAgB,KAAK,CAAC,YAAY,QAAQ,CAChE,CAAC;IACJ,CAAC;IAAC,OAAO,GAAQ,EAAE,CAAC;QAClB,OAAO,CAAC,KAAK,CAAC,qCAAqC,GAAG,CAAC,OAAO,EAAE,CAAC,CAAC;IACpE,CAAC;AACH,CAAC;AAED,SAAS,iBAAiB,CAAC,EAAW;IACpC,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IACvB,MAAM,UAAU,GAAG,IAAI,IAAI,CAAC,GAAG,GAAG,EAAE,GAAG,EAAE,GAAG,IAAI,CAAC,CAAC,WAAW,EAAE,CAAC;IAChE,MAAM,SAAS,GAAG,IAAI,IAAI,CAAC,GAAG,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,IAAI,CAAC,CAAC,WAAW,EAAE,CAAC;IAEpE,MAAM,GAAG,GAAG,EAAE,CAAC,QAAQ,CAAC;IAExB,MAAM,EAAE,GAAG,GAAG,CAAC,OAAO,CAAC;;;GAGtB,CAAC,CAAC,GAAG,CAAC,UAAU,CAAqC,CAAC;IAEvD,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC;;;GAGvB,CAAC,CAAC,GAAG,CAAC,SAAS,CAAqC,CAAC;IAEtD,OAAO;QACL,QAAQ,EAAE,EAAE,CAAC,KAAK;QAClB,aAAa,EAAE,EAAE,CAAC,KAAK;QACvB,OAAO,EAAE,GAAG,CAAC,KAAK;QAClB,YAAY,EAAE,GAAG,CAAC,KAAK;KACxB,CAAC;AACJ,CAAC"}

package/dist/engine/executor.d.ts

@@ -1,75 +0,0 @@
-/**
- * Configuration for the Agent SDK executor.
- */
-export interface ExecutorConfig {
-    /** Path to the MCP server config for GitHub tools */
-    mcpConfigPath: string;
-    /** Working directory for the agent */
-    cwd?: string;
-    /** Maximum conversation turns */
-    maxTurns?: number;
-    /** Model to use */
-    model?: string;
-    /** Path to agent context directory (soul.md, rules.md, etc.) */
-    agentContextDir?: string;
-    /** Directory for persistent state (sessions, logs). Mounted as Docker volume. */
-    stateDir?: string;
-    /** Directory for agent sandboxes (cloned repos). */
-    sandboxDir?: string;
-}
-/**
- * Result from an agent execution.
- */
-export interface ExecutionResult {
-    success: boolean;
-    output: string;
-    turns: number;
-    error?: string;
-    durationMs: number;
-    /** Session id captured from the agent's stream-json `system/init` line. */
-    sessionId?: string;
-    /** Total USD cost reported by Claude on the result message. */
-    costUsd?: number;
-    /** Tokens billed as fresh input (no cache hit). */
-    inputTokens?: number;
-    /** Tokens that triggered a cache write. */
-    cacheCreationInputTokens?: number;
-    /** Tokens served from prompt cache (heavily discounted). */
-    cacheReadInputTokens?: number;
-    /** Output tokens generated by the model. */
-    outputTokens?: number;
-    /** API-side duration (excludes orchestrator overhead). */
-    apiDurationMs?: number;
-    /** Stop subtype from the result message ("success" / "error_*" / etc.). */
-    stopReason?: string;
-}
-export type GitAccessProfile = "read" | "issues-write" | "review-write" | "repo-write";
-export interface GitSandboxAccess {
-    owner: string;
-    repo: string;
-    profile: GitAccessProfile;
-    /**
-     * When true, sandbox MCP can mint/refresh tokens using the app PEM.
-     * Leave false for lower-trust runs to keep private key material inaccessible.
-     */
-    allowMcpAppAuth?: boolean;
-}
-/**
- * Execute an agent task.
- *
- * Execution modes (automatic):
- * 1. Docker sandbox — if Docker + sandbox image available, runs `claude --print`
- *    inside an isolated container. Full sandboxing.
- * 2. Direct — runs Agent SDK query() in-process. Local dev fallback.
- */
-export declare function executeAgent(prompt: string, config: ExecutorConfig, opts?: {
-    taskId?: string;
-    /**
-     * Fired as soon as the agent's stream-json `system/init` line arrives,
-     * before the run completes. Used by the runner to persist the session id
-     * onto the in-flight DB row so the dashboard can show live logs.
-     */
-    onSessionId?: (sessionId: string) => void;
-    /** Optional per-run GitHub token policy for sandbox tooling. */
-    githubAccess?: GitSandboxAccess;
-}): Promise<ExecutionResult>;