tau-coding-agent 0.1.2 → 0.1.4

package/README.md

@@ -1,6 +1,6 @@
 # tau-coding-agent
 
-Coding-focused Tau package for [Pi](https://pi.dev).
+Coding agent built on top of [Pi](https://pi.dev).
 
 Tau turns Pi's minimal core into an opinionated, polished coding-agent setup with web search, code review workflows, sandboxing, local memory, usage reporting, and a curated skill set for day-to-day software work.
 
@@ -18,26 +18,27 @@ pi install -l npm:tau-coding-agent
 
 ## Extensions
 
-| Extension          | Command              | Description                                                                                                               |
-| ------------------ | -------------------- | ------------------------------------------------------------------------------------------------------------------------- |
-| `answer`           | `/answer`            | Extract and interactively answer agent questions.                                                                         |
-| `branch-term`      | `/branch`            | Open a new terminal on the current session's git branch.                                                                  |
-| `btw`              | `/btw`               | Run a one-off side request with read-only tools and no context persistence.                                               |
-| `ghostty`          | —                    | Ghostty tab title enhancements while the agent is working, waiting, or idle.                                              |
-| `git-diff-stats`   | —                    | Status bar diff stats for local changes in the current repo.                                                              |
-| `git-pr-status`    | —                    | Status bar PR number and link for the current branch.                                                                     |
-| `insights`         | `/insights`          | Analyze Pi sessions and suggest reusable instructions, templates, skills, and extensions.                                 |
-| `interlude`        | `alt+x`              | Stash the current message draft, send one interlude message, then restore the draft.                                      |
-| `loop`             | `/loop`              | Repeat a prompt until the agent signals success.                                                                          |
-| `memory`           | `/memory`            | Opt-in project-local memory for learning and continuity across sessions.                                                  |
-| `notify`           | —                    | Terminal notification when the agent is waiting for input.                                                                |
-| `openai-fast`      | `/fast`              | Toggle priority service tier for supported OpenAI models.                                                                 |
-| `openai-verbosity` | `/verbosity`         | Set verbosity for supported OpenAI models.                                                                                |
-| `review`           | `/review`, `/triage` | Multi-focus review and PR feedback triage for PRs, branches, commits, and local changes, with integrated follow-up fixes. |
-| `sandbox`          | `/sandbox`           | OS-level sandboxing for bash commands with runtime overrides.                                                             |
-| `usage`            | `/usage`             | Historical provider usage breakdown with all-provider history and live quota snapshots.                                   |
-| `websearch`        | —                    | Web search via Gemini, OpenAI, or Claude, leveraging Pi or browser session credentials.                                   |
-| `worktree`         | `/worktree`          | Create, list, and archive git worktrees, optionally opening them in a new terminal or tmux pane.                          |
+| Extension           | Command              | Description                                                                                                               |
+| ------------------- | -------------------- | ------------------------------------------------------------------------------------------------------------------------- |
+| `answer`            | `/answer`            | Extract and interactively answer agent questions.                                                                         |
+| `branch-term`       | `/branch`            | Open a new terminal on the current session's git branch.                                                                  |
+| `btw`               | `/btw`               | Run a one-off side request with read-only tools and no context persistence.                                               |
+| `ghostty`           | —                    | Ghostty tab title enhancements while the agent is working, waiting, or idle.                                              |
+| `git-diff-stats`    | —                    | Status bar diff stats for local changes in the current repo.                                                              |
+| `git-pr-status`     | —                    | Status bar PR number and link for the current branch.                                                                     |
+| `insights`          | `/insights`          | Analyze Pi sessions and suggest reusable instructions, templates, skills, and extensions.                                 |
+| `interlude`         | `alt+x`              | Stash the current message draft, send one interlude message, then restore the draft.                                      |
+| `loop`              | `/loop`              | Repeat a prompt until the agent signals success.                                                                          |
+| `memory`            | `/memory`            | Opt-in project-local memory for learning and continuity across sessions.                                                  |
+| `notify`            | —                    | Terminal notification when the agent is waiting for input.                                                                |
+| `openai-fast`       | `/fast`              | Toggle priority service tier for supported OpenAI models.                                                                 |
+| `openai-verbosity`  | `/verbosity`         | Set verbosity for supported OpenAI models.                                                                                |
+| `review`            | `/review`, `/triage` | Multi-focus review and PR feedback triage for PRs, branches, commits, and local changes, with integrated follow-up fixes. |
+| `sandbox`           | `/sandbox`           | OS-level sandboxing for bash commands with runtime overrides.                                                             |
+| `tool-display-mode` | `ctrl+o`             | Cycle tool output between Pi's default rendering, expanded output, and compact summaries.                                 |
+| `usage`             | `/usage`             | Historical provider usage breakdown with all-provider history and live quota snapshots.                                   |
+| `websearch`         | —                    | Web search via Gemini, OpenAI, or Claude, leveraging Pi or browser session credentials.                                   |
+| `worktree`          | `/worktree`          | Create, list, and archive git worktrees, optionally opening them in a new terminal or tmux pane.                          |
 
 ## Skills
 

package/extensions/answer.ts

@@ -10,9 +10,9 @@
  * 4. Submits the compiled answers when done
  */
 
-import { complete, type Model, type Api, type UserMessage } from "@mariozechner/pi-ai";
-import type { ExtensionAPI } from "@mariozechner/pi-coding-agent";
-import { BorderedLoader } from "@mariozechner/pi-coding-agent";
+import { complete, type Model, type Api, type UserMessage } from "@earendil-works/pi-ai";
+import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+import { BorderedLoader } from "@earendil-works/pi-coding-agent";
 import {
   type Component,
   Editor,
@@ -23,7 +23,7 @@ import {
   type TUI,
   visibleWidth,
   wrapTextWithAnsi,
-} from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-tui";
 
 const SYSTEM_PROMPT = `You extract items that need user input from assistant text:
 

package/extensions/branch-term.ts

@@ -2,14 +2,14 @@ import { spawn, spawnSync } from "node:child_process";
 import { randomUUID } from "node:crypto";
 import * as fs from "node:fs";
 import * as path from "node:path";
-import type { ExtensionAPI, ExtensionCommandContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionCommandContext } from "@earendil-works/pi-coding-agent";
 import {
   CURRENT_SESSION_VERSION,
   SessionManager,
   highlightCode,
   type SessionHeader,
-} from "@mariozechner/pi-coding-agent";
-import { Box, Text } from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-coding-agent";
+import { Box, Text } from "@earendil-works/pi-tui";
 
 const TERMINAL_FLAG = "branch-term";
 const TMUX_LAYOUT_FLAG = "branch-tmux-layout";

package/extensions/btw.ts

@@ -1,5 +1,5 @@
-import type { AgentMessage } from "@mariozechner/pi-agent-core";
-import type { AssistantMessage, ThinkingLevel } from "@mariozechner/pi-ai";
+import type { AgentMessage } from "@earendil-works/pi-agent-core";
+import type { AssistantMessage, ThinkingLevel } from "@earendil-works/pi-ai";
 import {
   DefaultResourceLoader,
   SessionManager,
@@ -11,7 +11,7 @@ import {
   type ExtensionCommandContext,
   type ExtensionContext,
   type Theme,
-} from "@mariozechner/pi-coding-agent";
+} from "@earendil-works/pi-coding-agent";
 import {
   Key,
   Markdown,
@@ -21,7 +21,7 @@ import {
   type TUI,
   visibleWidth,
   wrapTextWithAnsi,
-} from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-tui";
 
 const STATUS_KEY = "0-btw";
 const RESULT_MARKDOWN_THEME = getMarkdownTheme();

package/extensions/ghostty.ts

@@ -10,7 +10,7 @@
  * Pi now emits native OSC 9;4 progress indicators, so this extension only manages the title.
  */
 
-import type { ExtensionAPI, ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 import path from "node:path";
 
 const STATUS_SPINNER_INTERVAL_MS = 80;

package/extensions/git-diff-stats.ts

@@ -3,7 +3,7 @@ import {
   isWriteToolResult,
   type ExtensionAPI,
   type ExtensionContext,
-} from "@mariozechner/pi-coding-agent";
+} from "@earendil-works/pi-coding-agent";
 import { spawn } from "node:child_process";
 import { existsSync } from "node:fs";
 import { copyFile, mkdtemp, rm } from "node:fs/promises";

package/extensions/git-pr-status.ts

@@ -2,7 +2,7 @@ import {
   isBashToolResult,
   type ExtensionAPI,
   type ExtensionContext,
-} from "@mariozechner/pi-coding-agent";
+} from "@earendil-works/pi-coding-agent";
 
 const STATUS_KEY = "1-git-pr-status";
 

package/extensions/insights.ts

@@ -1,4 +1,4 @@
-import { complete, type Api, type Model, type UserMessage } from "@mariozechner/pi-ai";
+import { complete, type Api, type Model, type UserMessage } from "@earendil-works/pi-ai";
 import {
   BorderedLoader,
   SessionManager,
@@ -13,7 +13,7 @@ import {
   type SessionEntry,
   type SessionHeader,
   type Theme,
-} from "@mariozechner/pi-coding-agent";
+} from "@earendil-works/pi-coding-agent";
 import {
   Key,
   Markdown,
@@ -23,7 +23,7 @@ import {
   type TUI,
   visibleWidth,
   wrapTextWithAnsi,
-} from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-tui";
 import { createHash, randomBytes } from "node:crypto";
 import fs from "node:fs/promises";
 import path from "node:path";

package/extensions/interlude.ts

@@ -1,7 +1,7 @@
 import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
-import type { ExtensionAPI, ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 
 const DEFAULT_SHORTCUTS = ["alt+x"] as const;
 const STATUS_KEY = "interlude";

package/extensions/loop.ts

@@ -7,15 +7,15 @@
  */
 
 import { Type } from "typebox";
-import { complete, type Api, type Model, type UserMessage } from "@mariozechner/pi-ai";
+import { complete, type Api, type Model, type UserMessage } from "@earendil-works/pi-ai";
 import {
   compact,
   defineTool,
   DynamicBorder,
   type ExtensionAPI,
   type ExtensionContext,
-} from "@mariozechner/pi-coding-agent";
-import { Container, type SelectItem, SelectList, Text } from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-coding-agent";
+import { Container, type SelectItem, SelectList, Text } from "@earendil-works/pi-tui";
 
 const LOOP_PRESETS = [
   { value: "tests", label: "Until tests pass", description: "" },

package/extensions/memory.ts

@@ -1,11 +1,17 @@
-import { complete, StringEnum, type Api, type Model, type UserMessage } from "@mariozechner/pi-ai";
+import {
+  complete,
+  StringEnum,
+  type Api,
+  type Model,
+  type UserMessage,
+} from "@earendil-works/pi-ai";
 import {
   defineTool,
   withFileMutationQueue,
   type ExtensionAPI,
   type ExtensionCommandContext,
   type ExtensionContext,
-} from "@mariozechner/pi-coding-agent";
+} from "@earendil-works/pi-coding-agent";
 import { Type } from "typebox";
 import fs from "node:fs/promises";
 import path from "node:path";

package/extensions/notify.ts

@@ -10,7 +10,7 @@
  * - Windows toast: Windows Terminal (WSL)
  */
 
-import type { ExtensionAPI, ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 
 const REVIEW_EVENT_START = "review:start";
 const REVIEW_EVENT_END = "review:end";

package/extensions/openai-fast.ts

@@ -1,7 +1,7 @@
 import { mkdir, readFile, writeFile } from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import type { ExtensionAPI, ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 
 const STATUS_KEY = "openai-fast";
 const SUPPORTED_APIS = new Set([

package/extensions/openai-verbosity.ts

@@ -1,7 +1,7 @@
 import { mkdir, readFile, writeFile } from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import type { ExtensionAPI, ExtensionContext } from "@mariozechner/pi-coding-agent";
+import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 
 const STATUS_KEY = "openai-verbosity";
 const SUPPORTED_APIS = new Set([

package/extensions/review.ts

@@ -10,6 +10,7 @@
  *   when that message is a matching review payload, even if stale. Otherwise it
  *   runs a fresh review first. If that fresh review becomes stale while running,
  *   /fix shows the findings and waits for an explicit rerun before applying fixes.
+ *   context=... guides the fix pass without invalidating reusable review payloads.
  *
  * Key behavior:
  * - /review is findings-only (no direct edits).
@@ -37,9 +38,9 @@ import type {
   ExtensionAPI,
   ExtensionCommandContext,
   ExtensionContext,
-} from "@mariozechner/pi-coding-agent";
-import { supportsXhigh, type Api, type Model } from "@mariozechner/pi-ai";
-import { matchesKey } from "@mariozechner/pi-tui";
+} from "@earendil-works/pi-coding-agent";
+import { getSupportedThinkingLevels, type Api, type Model } from "@earendil-works/pi-ai";
+import { matchesKey } from "@earendil-works/pi-tui";
 import { spawn, type ChildProcess } from "node:child_process";
 import { createHash } from "node:crypto";
 import path from "node:path";
@@ -173,7 +174,7 @@ const REVIEW_FOCUSES: Record<FocusName, FocusDefinition> = {
   },
 };
 
-const REVIEW_ADDITIONAL_CONTEXT_SECTION_PROMPT = `Additional context from user:
+const ADDITIONAL_CONTEXT_SECTION_PROMPT = `Additional context from user:
 {ADDITIONAL_CONTEXT}
 `;
 
@@ -245,7 +246,7 @@ Output formatting requirements:
 - Do not include the pipe character in any cell text (including inside backticks). Avoid regex alternation patterns like (a|b); rewrite checks without pipes and separate multiple checks with semicolons.
 - Decision values must be exactly fixed or skipped.
 
-Review findings:
+{FIX_ADDITIONAL_CONTEXT_SECTION}Review findings:
 
 {REVIEW_FINDINGS_JSON}
 
@@ -430,10 +431,16 @@ type ParsedRequest = {
   target: ReviewTarget;
   mode: ReviewRequestMode;
   models: string[];
-  rawArgs: string;
+  targetExplicit: boolean;
   additionalContext?: string;
 };
 
+type RequestSignaturePayload = {
+  target: unknown;
+  models: string[];
+  additionalContext?: string | null;
+};
+
 type ReviewFingerprint = {
   headSha: string;
   branch: string;
@@ -878,7 +885,9 @@ Run findings-only code review in 4 parallel focuses (general, reuse, quality, ef
 
 ### Options
 - \`models=<a,b>\`: run all review focuses for each listed model.
-- \`context=<text>\`: add extra guidance to every review focus.
+- \`context=<text>\`: add extra guidance.
+  - For \`/review\`, guides every review focus.
+  - For \`/fix\`, guides the fix pass; if \`/fix\` must run a fresh review, it guides review too.
   - For spaces, quote the value, e.g. \`context="security and backpressure"\`.
 
 ### Examples
@@ -891,6 +900,7 @@ Run findings-only code review in 4 parallel focuses (general, reuse, quality, ef
 - \`/triage https://github.com/owner/repo/pull/123\`
 - \`/fix\`
 - \`/fix help\`
+- \`/fix context="do not test mocks"\`
 - \`/fix branch main models=sonnet\`
 
 ### /fix behavior
@@ -931,7 +941,12 @@ function parseKeyValueOption(
 function parseRequestArgs(args: string | undefined): ParsedRequest {
   const raw = args?.trim() ?? "";
   if (!raw) {
-    return { target: { type: "auto" }, mode: "auto", models: [], rawArgs: raw };
+    return {
+      target: { type: "auto" },
+      mode: "auto",
+      models: [],
+      targetExplicit: false,
+    };
   }
 
   const tokens = tokenizeArgs(raw);
@@ -984,16 +999,16 @@ function parseRequestArgs(args: string | undefined): ParsedRequest {
         return "custom";
     }
   };
-  const withMeta = (target: ReviewTarget): ParsedRequest => ({
+  const withMeta = (target: ReviewTarget, targetExplicit: boolean): ParsedRequest => ({
     target,
     mode: toMode(target),
     models,
-    rawArgs: raw,
+    targetExplicit,
     additionalContext,
   });
 
   if (modeTokens.length === 0) {
-    return withMeta({ type: "auto" });
+    return withMeta({ type: "auto" }, false);
   }
 
   const mode = modeTokens[0].toLowerCase();
@@ -1004,7 +1019,7 @@ function parseRequestArgs(args: string | undefined): ParsedRequest {
       throw new Error(
         `${mode} mode does not accept positional args. Use models=... and/or context=...`,
       );
-    return withMeta({ type: mode });
+    return withMeta({ type: mode }, true);
   }
 
   if (mode === "branch" || mode === "commit" || mode === "pr") {
@@ -1023,20 +1038,20 @@ function parseRequestArgs(args: string | undefined): ParsedRequest {
       );
     }
 
-    if (mode === "branch") return withMeta({ type: "branch", branch: rest[0] });
-    if (mode === "commit") return withMeta({ type: "commit", sha: rest[0] });
-    return withMeta({ type: "pr", ref: rest[0] });
+    if (mode === "branch") return withMeta({ type: "branch", branch: rest[0] }, true);
+    if (mode === "commit") return withMeta({ type: "commit", sha: rest[0] }, true);
+    return withMeta({ type: "pr", ref: rest[0] }, true);
   }
   if (mode === "folder") {
     if (rest.length === 0)
       throw new Error("folder mode requires at least one path (e.g. /review folder src docs)");
-    return withMeta({ type: "folder", paths: rest.map((p) => p.trim()).filter(Boolean) });
+    return withMeta({ type: "folder", paths: rest.map((p) => p.trim()).filter(Boolean) }, true);
   }
   if (mode === "custom") {
     const instructions = rest.join(" ").trim();
     if (!instructions)
       throw new Error('custom mode requires instructions (e.g. /review custom "focus on auth")');
-    return withMeta({ type: "custom", instructions });
+    return withMeta({ type: "custom", instructions }, true);
   }
 
   throw new Error(
@@ -1044,7 +1059,14 @@ function parseRequestArgs(args: string | undefined): ParsedRequest {
   );
 }
 
-function buildRequestSignature(request: ParsedRequest): string {
+function normalizeRequestModels(models: string[]): string[] {
+  return Array.from(new Set(models)).sort((a, b) => a.localeCompare(b));
+}
+
+function buildRequestSignaturePayload(
+  request: ParsedRequest,
+  options: { includeAdditionalContext?: boolean } = {},
+): RequestSignaturePayload {
   const additionalContext = request.additionalContext?.trim();
   const target =
     request.target.type === "folder"
@@ -1062,15 +1084,55 @@ function buildRequestSignature(request: ParsedRequest): string {
               ? { type: "commit", sha: request.target.sha }
               : { type: request.target.type };
 
-  return JSON.stringify({
+  const payload: RequestSignaturePayload = {
     target,
-    models: Array.from(new Set(request.models)).sort((a, b) => a.localeCompare(b)),
-    additionalContext: additionalContext && additionalContext.length > 0 ? additionalContext : null,
-  });
+    models: normalizeRequestModels(request.models),
+  };
+  if (options.includeAdditionalContext ?? true) {
+    payload.additionalContext =
+      additionalContext && additionalContext.length > 0 ? additionalContext : null;
+  }
+  return payload;
 }
 
-function reviewMatchesRequest(details: ReviewMessageDetails, request: ParsedRequest): boolean {
-  return details.request.signature === buildRequestSignature(request);
+function buildRequestSignature(request: ParsedRequest): string {
+  return JSON.stringify(buildRequestSignaturePayload(request));
+}
+
+function buildContextlessRequestSignature(request: ParsedRequest): string {
+  return JSON.stringify(buildRequestSignaturePayload(request, { includeAdditionalContext: false }));
+}
+
+function parseRequestSignaturePayload(signature: string): RequestSignaturePayload | null {
+  try {
+    const payload: unknown = JSON.parse(signature);
+    if (!payload || typeof payload !== "object") return null;
+    const { target, models } = payload as { target?: unknown; models?: unknown };
+    if (target === undefined || !Array.isArray(models)) return null;
+    const parsedModels = models.filter((model): model is string => typeof model === "string");
+    if (parsedModels.length !== models.length) return null;
+    return { target, models: normalizeRequestModels(parsedModels) };
+  } catch {
+    return null;
+  }
+}
+
+function buildContextlessStoredRequestSignature(signature: string): string | null {
+  const payload = parseRequestSignaturePayload(signature);
+  if (!payload) return null;
+  return JSON.stringify({ target: payload.target, models: payload.models });
+}
+
+function isOpenEndedFixRequest(request: ParsedRequest): boolean {
+  return !request.targetExplicit && request.models.length === 0;
+}
+
+function reviewMatchesFixRequest(details: ReviewMessageDetails, request: ParsedRequest): boolean {
+  if (isOpenEndedFixRequest(request)) return true;
+  return (
+    buildContextlessStoredRequestSignature(details.request.signature) ===
+    buildContextlessRequestSignature(request)
+  );
 }
 
 // --- Git & fingerprinting ---
@@ -1794,17 +1856,19 @@ function buildScopeInstructions(scope: ResolvedScope): string {
   }
 }
 
+function buildAdditionalContextSection(additionalContext: string | undefined): string {
+  const trimmed = additionalContext?.trim();
+  if (!trimmed) return "";
+  return ADDITIONAL_CONTEXT_SECTION_PROMPT.replace("{ADDITIONAL_CONTEXT}", () => trimmed);
+}
+
 function buildFocusPrompt(
   focus: FocusName,
   scopeInstructions: string,
   projectGuidelines: string | null,
   additionalContext: string | undefined,
 ): string {
-  const additionalContextSection = additionalContext?.trim()
-    ? REVIEW_ADDITIONAL_CONTEXT_SECTION_PROMPT.replace("{ADDITIONAL_CONTEXT}", () =>
-        additionalContext.trim(),
-      )
-    : "";
+  const additionalContextSection = buildAdditionalContextSection(additionalContext);
   const projectGuidelinesSection = projectGuidelines
     ? REVIEW_PROJECT_GUIDELINES_SECTION_PROMPT.replace(
         "{PROJECT_GUIDELINES}",
@@ -3225,7 +3289,7 @@ function createResolvedReviewProviderCandidateFromModel(
   return createResolvedReviewProviderCandidate({
     baseModelArg: `${model.provider}/${model.id}`,
     supportsThinking: model.reasoning,
-    supportsXhigh: supportsXhigh(model),
+    supportsXhigh: getSupportedThinkingLevels(model).includes("xhigh"),
   });
 }
 
@@ -3353,7 +3417,9 @@ async function resolveModels(
         createResolvedReviewProviderCandidate({
           baseModelArg,
           supportsThinking: ctx.model?.reasoning,
-          supportsXhigh: ctx.model ? supportsXhigh(ctx.model) : undefined,
+          supportsXhigh: ctx.model
+            ? getSupportedThinkingLevels(ctx.model).includes("xhigh")
+            : undefined,
         }),
       ],
     }),
@@ -4083,7 +4149,10 @@ async function runTriagePipeline(
 
 // --- Command handlers ---
 
-function buildFixPrompt(reviewMessageDetails: ReviewMessageDetails): string {
+function buildFixPrompt(
+  reviewMessageDetails: ReviewMessageDetails,
+  additionalContext: string | undefined,
+): string {
   const worklistPayload = JSON.stringify(
     {
       scope: reviewMessageDetails.scope,
@@ -4101,8 +4170,12 @@ function buildFixPrompt(reviewMessageDetails: ReviewMessageDetails): string {
     null,
     2,
   );
+  const additionalContextSection = buildAdditionalContextSection(additionalContext);
 
-  return FIX_PROMPT.replace("{REVIEW_FINDINGS_JSON}", () => worklistPayload);
+  return FIX_PROMPT.replace(
+    "{FIX_ADDITIONAL_CONTEXT_SECTION}",
+    () => additionalContextSection,
+  ).replace("{REVIEW_FINDINGS_JSON}", () => worklistPayload);
 }
 
 function parseTriagePrRef(
@@ -4259,7 +4332,7 @@ export default function reviewExtension(pi: ExtensionAPI) {
 
   pi.registerCommand("fix", {
     description:
-      "Fix findings when the last message is a matching /review result. Otherwise, runs review first. If the last review is stale, /fix warns and continues; if a /fix-started review goes stale mid-run, /fix shows the findings and applies no fixes. Use /review help for modes/options.",
+      "Fix findings when the last message is a matching /review result. Otherwise, runs review first. context=... guides the fix pass without forcing a fresh review. If the last review is stale, /fix warns and continues; if a /fix-started review goes stale mid-run, /fix shows the findings and applies no fixes. Use /review help for modes/options.",
     getArgumentCompletions: getReviewArgumentCompletions,
     handler: async (args, ctx) => {
       const request = parseCommandRequest(pi, args, ctx);
@@ -4273,9 +4346,9 @@ export default function reviewExtension(pi: ExtensionAPI) {
 
       try {
         let reviewDetails = getLastMessageReviewDetails(ctx);
-        const lastReviewMatchesRequest =
-          !request.rawArgs ||
-          (reviewDetails ? reviewMatchesRequest(reviewDetails, request) : false);
+        const lastReviewMatchesRequest = reviewDetails
+          ? reviewMatchesFixRequest(reviewDetails, request)
+          : false;
         const shouldRunFreshReview = !reviewDetails || !lastReviewMatchesRequest;
 
         if (reviewDetails && !shouldRunFreshReview) {
@@ -4332,7 +4405,7 @@ export default function reviewExtension(pi: ExtensionAPI) {
           return;
         }
 
-        const fixPrompt = buildFixPrompt(reviewDetails);
+        const fixPrompt = buildFixPrompt(reviewDetails, request.additionalContext);
         if (ctx.isIdle()) {
           pi.sendUserMessage(fixPrompt);
         } else {