peaks-loop 4.0.25 → 4.0.27

package/dist/services/retrospective/retrospective-search-service.js

@@ -73,35 +73,3 @@ export function searchRetrospective(input) {
         };
     });
 }
-import { compressPrompt as compressPromptForRetro } from '../context/headroom-client.js';
-import { loadPreferences as loadPreferencesForRetro } from '../preferences/preferences-service.js';
-import { shouldCompressResults as shouldCompressResultsForRetro } from '../context/headroom-prefs.js';
-export async function searchRetrospectiveWithResults(input, options = {}) {
-    const matches = searchRetrospective(input);
-    if (options.compressResults !== true) {
-        return { matches, compressedResults: null };
-    }
-    const projectRoot = input.projectRoot ?? process.cwd();
-    const prefs = loadPreferencesForRetro(projectRoot).headroom;
-    const joinedText = matches.map((m) => `${m.title}: ${m.summary}`).join('\n');
-    const joinedBytes = Buffer.byteLength(joinedText, 'utf8');
-    const decision = shouldCompressResultsForRetro(prefs, joinedBytes, 'retrospectiveSearch');
-    if (!decision.compress) {
-        return { matches, compressedResults: null };
-    }
-    const result = await compressPromptForRetro(joinedText, decision.mode);
-    if (result.warning !== null || result.compressedPrompt === null) {
-        return { matches, compressedResults: null };
-    }
-    return {
-        matches,
-        compressedResults: {
-            mode: decision.mode,
-            originalSize: result.originalSize,
-            compressedSize: result.compressedSize,
-            compressionRatio: result.compressionRatio,
-            compressedText: result.compressedPrompt,
-            warning: result.warning
-        }
-    };
-}

package/dist/services/session/binding-status-service.d.ts

@@ -38,6 +38,17 @@ export type BindingStatusView = {
  * helper that does no filesystem writes; safe to call from any test.
  */
 export declare function loadBindingStatus(projectRoot: string): BindingStatusView;
+export declare function readOuterSessionId(): string;
+/**
+ * Resolve the outer (harness / IDE) session id for the auto-compact
+ * context-percent probe. Order: env signal first (PEAKS_OUTER_SESSION_ID /
+ * CLAUDE_CODE_SESSION_ID), then the bound peaks session's recorded
+ * `outerSessionId` (`session.json` meta). Returns `undefined` when neither is
+ * available — the caller passes `undefined` through and the adapter's fallback
+ * returns null → conservative-fallback. Vendor-neutral: this helper has no
+ * IDE-specific naming; the env-var read lives alongside the binding layer.
+ */
+export declare function resolveOuterSessionId(projectRoot: string, sessionId: string, env?: NodeJS.ProcessEnv): string | undefined;
 /**
  * Render the binding as a pipeable ASCII table. The columns are fixed
  * (no truncation, no wrapping) so a downstream `awk` / `cut` pipeline

package/dist/services/session/binding-status-service.js

@@ -22,6 +22,7 @@
 import { existsSync } from 'node:fs';
 import { join } from 'node:path';
 import { readBinding } from './binding-store.js';
+import { getSessionMeta } from './session-manager.js';
 /**
  * Read the binding from disk and assemble the read-only view. Pure
  * helper that does no filesystem writes; safe to call from any test.
@@ -37,14 +38,34 @@ export function loadBindingStatus(projectRoot) {
         : !Object.values(binding.instances).some((inst) => inst.callerId.startsWith(outerSessionId));
     return { binding, source, projectRoot, stale, outerSessionId };
 }
-function readOuterSessionId() {
-    const peaks = process.env.PEAKS_OUTER_SESSION_ID;
+export function readOuterSessionId() {
+    return readOuterSessionIdFromEnv(process.env) ?? 'unknown';
+}
+/** Pure env read (no process.env dependency) so callers can inject an env. */
+function readOuterSessionIdFromEnv(env) {
+    const peaks = env.PEAKS_OUTER_SESSION_ID;
     if (typeof peaks === 'string' && peaks.length > 0)
         return peaks;
-    const claude = process.env.CLAUDE_CODE_SESSION_ID;
+    const claude = env.CLAUDE_CODE_SESSION_ID;
     if (typeof claude === 'string' && claude.length > 0)
         return claude;
-    return 'unknown';
+    return undefined;
+}
+/**
+ * Resolve the outer (harness / IDE) session id for the auto-compact
+ * context-percent probe. Order: env signal first (PEAKS_OUTER_SESSION_ID /
+ * CLAUDE_CODE_SESSION_ID), then the bound peaks session's recorded
+ * `outerSessionId` (`session.json` meta). Returns `undefined` when neither is
+ * available — the caller passes `undefined` through and the adapter's fallback
+ * returns null → conservative-fallback. Vendor-neutral: this helper has no
+ * IDE-specific naming; the env-var read lives alongside the binding layer.
+ */
+export function resolveOuterSessionId(projectRoot, sessionId, env = process.env) {
+    const envOuter = readOuterSessionIdFromEnv(env);
+    if (envOuter !== undefined)
+        return envOuter;
+    const recorded = getSessionMeta(projectRoot, sessionId)?.outerSessionId;
+    return typeof recorded === 'string' && recorded.length > 0 ? recorded : undefined;
 }
 /**
  * Render the binding as a pipeable ASCII table. The columns are fixed

package/dist/services/skill/skill-search-service.d.ts

@@ -32,9 +32,9 @@ export declare const SkillSearchInputSchema: z.ZodObject<{
         status: "status";
         resume: "resume";
         audit: "audit";
+        ide: "ide";
         content: "content";
         test: "test";
-        ide: "ide";
         research: "research";
         doctor: "doctor";
         triage: "triage";

package/dist/services/slice/slice-benchmark-service.d.ts

@@ -1,5 +1,5 @@
 /**
- * Slice benchmark wrapper (slice 2026-06-14-fzf-headroom-rollout).
+ * Slice benchmark wrapper (slice 2026-06-14-fzf-rollout).
  *
  * Wraps `decomposeSlices` with a thin observability layer that records
  * total wall time, codegraph call count, confidence distribution, and

package/dist/services/slice/slice-benchmark-service.js

@@ -1,5 +1,5 @@
 /**
- * Slice benchmark wrapper (slice 2026-06-14-fzf-headroom-rollout).
+ * Slice benchmark wrapper (slice 2026-06-14-fzf-rollout).
  *
  * Wraps `decomposeSlices` with a thin observability layer that records
  * total wall time, codegraph call count, confidence distribution, and

package/dist/services/slice/slice-pick-service.d.ts

@@ -1,7 +1,7 @@
 /**
  * Slice Pick Service — thin wrapper around the generic fzf picker.
  *
- * v2 (slice 2026-06-14-fzf-headroom-rollout): the actual fzf
+ * v2 (slice 2026-06-14-fzf-rollout): the actual fzf
  * integration is now in `src/services/fuzzy-matching/fzf-pick-service.ts`.
  * This file only knows how to format / parse SliceCandidate rows and
  * pick the output path. The algorithm itself (slice-decompose-service)

package/dist/services/slice/slice-pick-service.js

@@ -1,7 +1,7 @@
 /**
  * Slice Pick Service — thin wrapper around the generic fzf picker.
  *
- * v2 (slice 2026-06-14-fzf-headroom-rollout): the actual fzf
+ * v2 (slice 2026-06-14-fzf-rollout): the actual fzf
  * integration is now in `src/services/fuzzy-matching/fzf-pick-service.ts`.
  * This file only knows how to format / parse SliceCandidate rows and
  * pick the output path. The algorithm itself (slice-decompose-service)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "peaks-loop",
-  "version": "4.0.25",
+  "version": "4.0.27",
   "description": "Loop Engineering CLI — workflow primitive / loop guards / evaluators / slice orchestration",
   "author": "SquabbyZ",
   "keywords": [
@@ -99,13 +99,12 @@
     "better-sqlite3": "^12.11.1",
     "commander": "^12.1.0",
     "fzf": "^0.5.2",
-    "headroom-ai": "0.22.4",
     "yaml": "^2.9.0",
     "zod": "^4.4.3",
-    "peaks-loop-internal-runtime": "0.0.10",
-    "peaks-loop-mut": "0.1.23",
-    "peaks-loop-shared-channel": "0.0.27",
-    "peaks-loop-shared": "0.0.59"
+    "peaks-loop-mut": "0.1.25",
+    "peaks-loop-internal-runtime": "0.0.12",
+    "peaks-loop-shared": "0.0.61",
+    "peaks-loop-shared-channel": "0.0.29"
   },
   "devDependencies": {
     "@changesets/cli": "2.31.1",

package/skills/bee/peaks-qa/references/qa-context-governance.md

@@ -21,4 +21,4 @@ PROTOCOL (mandatory):
 
 ## G9 — QA prompt size self-check
 
-Same as RD: 50% soft warn, 75% `CONTEXT_NEAR_LIMIT`, 80% hard reject unless `--force`. QA test plans can grow large; prefer `--use-headroom balanced` for plans > 75%.
+Same as RD: 50% soft warn, 75% `CONTEXT_NEAR_LIMIT`, 80% hard reject unless `--force`. QA test plans can grow large; trim or split into multiple dispatches for plans > 75%.

package/skills/bee/peaks-rd/SKILL.md

@@ -235,9 +235,9 @@ The `karpathy-reviewer` sub-agent now reports its own runtime cost in the JSON e
 
 The main RD loop MUST call `peaks job karpathy-cost-check` after every `peaks request transition --state qa-handoff` and before the next-slice work begins. If the decision kind is `downgraded`, the LLM MUST honor the downgraded `warn` and proceed to the next slice; the `'block'` was an artifact of the reviewer's own cost, not the slice's quality. If the decision kind is `reported` (costRatio > 50, gate not `'block'`), the LLM MAY continue; the sediment will be appended by `peaks memory extract` at handoff time. The full design lives in `.peaks/memory/2026-07-30-karpathy-evaluation-cost-self-review-design.md` (sediment locked 2026-07-30).
 
-## Sub-agent context governance (G7 + G7.7 + G8 + G9 — slice #010)
+## Sub-agent context governance (G7 + G8 + G9 — slice #010)
 
-RD sub-agent prompt template MUST include the G7 path convention + G8.6 share protocol. Detailed protocol: `skills/peaks-code/references/context-governance.md` + `skills/peaks-code/references/headroom-integration.md`.
+RD sub-agent prompt template MUST include the G7 path convention + G8.6 share protocol. Detailed protocol: `skills/peaks-code/references/context-governance.md`.
 
 → see `references/rd-context-governance.md` for the full G7 / G8.6 / G9 protocol + RD sub-agent prompt template.
 

package/skills/bee/peaks-rd/references/rd-context-governance.md

@@ -31,6 +31,6 @@ PROTOCOL (mandatory):
 
 Before dispatching a sub-agent, RD self-checks prompt size:
 - < 50%: pass through.
-- 50-75%: soft warn (consider `--use-headroom`).
-- 75-80%: soft warn + `warnings: ["CONTEXT_NEAR_LIMIT"]` (mandatory suggest `--use-headroom`).
+- 50-75%: soft warn (consider trimming the prompt).
+- 75-80%: soft warn + `warnings: ["CONTEXT_NEAR_LIMIT"]` (trim or split into multiple dispatches).
 - 80%+: reject (CLI 兜底 returns `code: "PROMPT_TOO_LARGE"`). Use `--force` at CLI only when overriding; hook layer will still reject (RL-30).

package/skills/bee/peaks-txt/SKILL.md

@@ -305,7 +305,7 @@ The TXT handoff summarizes the slice by listing the artifact metas + the share e
 
 ### G9 — TXT prompt size self-check
 
-Same G9 threshold table. TXT handoff messages can grow large when the slice has many batches; use `--use-headroom` proactively for handoff prompts > 50%.
+Same G9 threshold table. TXT handoff messages can grow large when the slice has many batches; trim or split the handoff prompt proactively for prompts > 50%.
 
 
 ## L2 surface reference (post rid-l2-extended)

package/skills/bee/peaks-ui/SKILL.md

@@ -361,7 +361,7 @@ PROTOCOL (mandatory):
 
 ### G9 — UI prompt size self-check
 
-Same as RD/QA. Use `--use-headroom` proactively.
+Same as RD/QA. Trim or split the prompt proactively.
 
 
 ## L2 surface reference (post rid-l2-extended)

package/skills/peaks-code/SKILL.md

@@ -302,7 +302,7 @@ After final validation, refresh project-local standards via `peaks standards ini
 
 `peaks codegraph affected` is an optional project-analysis enhancement (untrusted supporting evidence) for role handoff. Code must not treat codegraph output as approval for scope, design, or QA verdict. Never mutate agent settings / hooks from codegraph; do not commit `.codegraph/` artifacts. RD writes `.peaks/_runtime/<sessionId>/rd/codegraph-context.md`; QA / TXT consume the same envelope. → `references/codegraph-orchestration.md`.
 
-## Sub-agent context governance (G7 + G7.7 + G8 + G9 — slice #010)
+## Sub-agent context governance (G7 + G8 + G9 — slice #010)
 
 Main LLM reducer sees metadata-only view (~200 chars/sub-agent); on-demand `Read` for full content. Threshold table: 50% soft warn, 75% `CONTEXT_NEAR_LIMIT`, 80% hard reject (CLI + hook double-guard). → `references/context-governance.md`.
 

package/skills/peaks-code/references/context-governance.md

@@ -1,7 +1,7 @@
-# Context Governance — G7 + G7.7 + G8 + G9 protocol details
+# Context Governance — G7 + G8 + G9 protocol details
 
-> Slice #010 (G7 + G7.7 + G8 + G9 context-governance push).
-> See: `.peaks/memory/sub-agent-context-minimal-occupation.md` + `sub-agent-shared-channel-cross-completion.md` + `sub-agent-headroom-forced-compression-gate.md` for the red lines.
+> Slice #010 (G7 + G8 + G9 context-governance push).
+> See: `.peaks/memory/sub-agent-context-minimal-occupation.md` + `sub-agent-shared-channel-cross-completion.md` for the red lines.
 
 ## G7 — sub-agent context minimal-occupation (metadata-only + 按需 Read)
 
@@ -56,26 +56,6 @@ On completion:
 
 3000-5000× improvement. Main LLM full-slice context net increase: < 10KB for 5 batches × 6 sub-agents.
 
-## G7.7 — headroom-ai integration (opt-in)
-
-### `--use-headroom` flag
-
-Opt-in flag on `peaks sub-agent dispatch`. Default `false` (G7 metadata-only remains the default).
-
-### Mode table
-
-| Mode | tokenBudget | Use case |
-|---|---|---|
-| `balanced` (default) | promptSize * 0.40 / 4 | General sub-agent dispatch |
-| `aggressive` | promptSize * 0.20 / 4 | Last-resort large prompt |
-| `conservative` | promptSize * 0.70 / 4 | Sensitive code analysis |
-
-### Failure mode (RL-22d / RL-32)
-
-- headroom daemon dead / proxy unreachable / times out
-- → `code: "HEADROOM_UNAVAILABLE"` warning + G7 metadata-only fallback
-- → NOT blocking (warn, then continue dispatch)
-
 ## G8 — cross sub-agent shared channel
 
 ### Path convention
@@ -132,8 +112,8 @@ PROTOCOL (mandatory):
 
 | Threshold | Prompt size | Behavior |
 |---|---|---|
-| 50% (early warn) | ≥ 128KB | Soft warning, suggest `--use-headroom` |
-| **75% (user red line)** | ≥ 192KB | Soft warn + mandatory suggest `--use-headroom`; `warnings: ["CONTEXT_NEAR_LIMIT"]` |
+| 50% (early warn) | ≥ 128KB | Soft warning, suggest trimming the prompt |
+| **75% (user red line)** | ≥ 192KB | Soft warn + mandatory trim/split suggestion; `warnings: ["CONTEXT_NEAR_LIMIT"]` |
 | **80% (hard reject)** | ≥ 204KB | Hard reject `code: "PROMPT_TOO_LARGE"`; `--force` allowed at CLI |
 | **90% (emergency)** | ≥ 230KB | Hard reject `code: "PROMPT_EMERGENCY"`; `--force` STILL rejects at 90% (no override) |
 
@@ -149,7 +129,7 @@ PROTOCOL (mandatory):
 
 ## AC mapping
 
-- AC-38..AC-43 (G7) + AC-44..AC-46 (G7.7) + AC-47..AC-49 (G8) + AC-50..AC-65 (G9)
+- AC-38..AC-43 (G7) + AC-47..AC-49 (G8) + AC-50..AC-65 (G9)
 - See PRD §Acceptance criteria.
 
 ---
@@ -171,13 +151,6 @@ Main LLM view format (G7.4.e):
 - qa-perf → .../artifacts/003-qa-perf-001.md (5KB, sha256:ghi789) summary: "p95 latency target ≤ 200ms"
 ```
 
-### G7.7 — headroom-ai integration (opt-in compression)
-
-> Body of `### G7.7`. If a sub-agent prompt is too large even after G7 metadata-only (e.g. 1MB artifact description, 5MB mid-prompt analysis), use `--use-headroom`:
-- Default `false` (G7 remains default).
-- Modes: `balanced` (default) | `aggressive` | `conservative`.
-- Failure: `HEADROOM_UNAVAILABLE` warning + G7 metadata-only fallback (NOT blocking).
-
 ### G8 — cross sub-agent shared channel (dispatcher-mediated indirect signal)
 
 > Body of `### G8`. Sub-agent A's completion **immediately** writes a shared entry; sub-agent B (still in flight) can read shared entries from sibling sub-agents. **This is NOT peer-to-peer messaging.** The dispatcher stores, the sub-agents read/write; A and B never directly talk.
@@ -192,7 +165,7 @@ Main LLM view format (G7.4.e):
 
 | Threshold | Prompt size | Behavior |
 |---|---|---|
-| 50% (early warn) | ≥ 128KB | Soft warning, suggest `--use-headroom` |
+| 50% (early warn) | ≥ 128KB | Soft warning, suggest trimming the prompt |
 | **75% (user red line)** | ≥ 192KB | Soft warn + `warnings: ["CONTEXT_NEAR_LIMIT"]` |
 | **80% (hard reject)** | ≥ 204KB | Hard reject `code: "PROMPT_TOO_LARGE"`; `--force` allowed at CLI |
 | 90% (emergency) | ≥ 230KB | Hard reject + `contextWarning: 'high'` |

package/skills/peaks-code/references/envelope-contract.md

@@ -33,19 +33,14 @@ Top-level fields (post-2.1.0):
 - `envelopeVersion: '2.1.0'`
 - `role: string` — sub-agent role string
 - `ide: string` — detected IDE label (`claude-code` etc.)
-- `originalPromptSize: number` — bytes of the un-compressed prompt
-- `promptSize: number` — bytes after headroom compression (or
-  equal to `originalPromptSize` when `--use-headroom` is off or
-  unavailable)
+- `originalPromptSize: number` — bytes of the caller-supplied prompt
+- `promptSize: number` — bytes of the composed dispatch prompt
 - `toolCall: { name: string; args: Record<string, unknown> }` —
   the per-IDE tool-call descriptor the LLM must execute
 - `dispatchRecordPath: string` — absolute path to the dispatch
   record on disk
 - `batchId: string` — uuid-like opaque token grouping one batch
 - `dispatchedInBatch: number` — current count after this dispatch
-- `headroomCompressed: boolean` — true if headroom-ai actually
-  reduced the prompt
-- `headroomResult: { mode, compressed, compressionRatio, tokensSaved, warning } | null`
 - `forcedAt: string | null` — ISO8601 when `--force` overrode
   the G9 hard-reject tier
 - `contextImpact: { promptBytes, artifactBytes, totalBytes }`

package/skills/peaks-code/references/sub-agent-dispatch.md

@@ -51,8 +51,8 @@ peaks sub-agent dispatch <role> --prompt <text> [--request-id <rid>] [--session-
 > scrollback. The dispatch record on disk (gitignored under
 > `.peaks/_sub_agents/`) keeps the prompt for the sub-agent to read;
 > CLI stdout stays metadata-only. Surface `originalPromptSize` +
-> `promptSize` so the LLM-side runner can still reason about headroom
-> without seeing the content. The prompt content the LLM should pass
+> `promptSize` so the LLM-side runner can still reason about the size
+> delta without seeing the content. The prompt content the LLM should pass
 > through lives in `data.toolCall.args.prompt` (the IDE-arg the
 > consumer passes to the tool), NOT in the outer envelope.
 
@@ -77,8 +77,6 @@ peaks sub-agent dispatch <role> --prompt <text> [--request-id <rid>] [--session-
     "dispatchRecordPath": ".peaks/_sub_agents/2026-06-06-session-5b1095/dispatch-002-2026-06-07-...-...json",
     "batchId": "<uuid>",
     "dispatchedInBatch": 3,
-    "headroomCompressed": false,
-    "headroomResult": null,
     "forcedAt": null,
     "contextImpact": {
       "promptBytes": 4321,

package/dist/services/context/headroom-client.d.ts

@@ -1,34 +0,0 @@
-export type HeadroomMode = 'balanced' | 'aggressive' | 'conservative';
-export interface HeadroomResult {
-    readonly compressed: boolean;
-    readonly originalSize: number;
-    readonly compressedSize: number;
-    readonly compressionRatio: number;
-    readonly mode: HeadroomMode;
-    /** `'HEADROOM_UNAVAILABLE'` on fallback; `null` on success. */
-    readonly warning: string | null;
-    /** Compressed prompt body. `null` if no compression happened. */
-    readonly compressedPrompt: string | null;
-    /** Tokens saved (from the SDK). 0 on fallback. */
-    readonly tokensSaved: number;
-}
-/**
- * Compress a prompt via headroom-ai. The `fallback: true` option is
- * non-negotiable: if the proxy daemon is unavailable, the SDK returns
- * `result.compressed = false` and the original messages; we surface
- * that as `HEADROOM_UNAVAILABLE` warning + G7 metadata-only fallback.
- */
-export declare function compressPrompt(prompt: string, mode?: HeadroomMode): Promise<HeadroomResult>;
-/**
- * Bridge interface: when `--use-headroom` is set, share entries written
- * via `peaks sub-agent share` MAY also flow through headroom's
- * `SharedContext`. Slice #010 implements a peak-internal shared channel
- * (see `shared-channel.ts`); the headroom-side `SharedContext` is a
- * separate concept that future slices can layer on. For now this
- * function is a stub that returns the peak-internal channel ID, which
- * is enough to demonstrate the bridge contract.
- */
-export declare function buildSharedContextBridge(batchId: string): {
-    peakChannelId: string;
-    headroomContextId: string;
-};

package/dist/services/context/headroom-client.js

@@ -1,117 +0,0 @@
-const DEFAULT_TIMEOUT_MS = 30_000;
-const DEFAULT_MODEL = 'claude-sonnet-4-5-20250929';
-// Approximate 1 token = 4 bytes for English text. This is a rough
-// heuristic; the SDK does its own tokenization internally.
-const BYTES_PER_TOKEN = 4;
-/**
- * Compress a prompt via headroom-ai. The `fallback: true` option is
- * non-negotiable: if the proxy daemon is unavailable, the SDK returns
- * `result.compressed = false` and the original messages; we surface
- * that as `HEADROOM_UNAVAILABLE` warning + G7 metadata-only fallback.
- */
-export async function compressPrompt(prompt, mode = 'balanced') {
-    const originalSize = Buffer.byteLength(prompt, 'utf8');
-    let compressFn = null;
-    try {
-        const mod = await import('headroom-ai');
-        if (typeof mod.compress === 'function') {
-            compressFn = mod.compress;
-        }
-    }
-    catch {
-        return fallback(originalSize, mode);
-    }
-    if (compressFn === null) {
-        return fallback(originalSize, mode);
-    }
-    const messages = [
-        { role: 'user', content: prompt }
-    ];
-    const opts = {
-        model: DEFAULT_MODEL,
-        timeout: DEFAULT_TIMEOUT_MS,
-        fallback: true, // CRITICAL: return original messages if proxy is down
-        retries: 1
-    };
-    if (mode === 'aggressive') {
-        opts.tokenBudget = Math.max(1, Math.floor(originalSize * 0.20 / BYTES_PER_TOKEN));
-    }
-    else if (mode === 'conservative') {
-        opts.tokenBudget = Math.max(1, Math.floor(originalSize * 0.70 / BYTES_PER_TOKEN));
-    }
-    else {
-        // balanced: target ~60% reduction
-        opts.tokenBudget = Math.max(1, Math.floor(originalSize * 0.40 / BYTES_PER_TOKEN));
-    }
-    let result;
-    try {
-        result = await compressFn(messages, opts);
-    }
-    catch {
-        return fallback(originalSize, mode);
-    }
-    if (result.compressed === false) {
-        return {
-            compressed: false,
-            originalSize,
-            compressedSize: originalSize,
-            compressionRatio: 1.0,
-            mode,
-            warning: 'HEADROOM_UNAVAILABLE',
-            compressedPrompt: null,
-            tokensSaved: 0
-        };
-    }
-    const compressedContent = extractContent(result.messages);
-    if (compressedContent === null) {
-        return fallback(originalSize, mode);
-    }
-    const compressedSize = Buffer.byteLength(compressedContent, 'utf8');
-    return {
-        compressed: true,
-        originalSize,
-        compressedSize,
-        compressionRatio: compressedSize / originalSize,
-        mode,
-        warning: null,
-        compressedPrompt: compressedContent,
-        tokensSaved: result.tokensSaved ?? 0
-    };
-}
-function fallback(originalSize, mode) {
-    return {
-        compressed: false,
-        originalSize,
-        compressedSize: originalSize,
-        compressionRatio: 1.0,
-        mode,
-        warning: 'HEADROOM_UNAVAILABLE',
-        compressedPrompt: null,
-        tokensSaved: 0
-    };
-}
-function extractContent(messages) {
-    if (!Array.isArray(messages) || messages.length === 0) {
-        return null;
-    }
-    const last = messages[messages.length - 1];
-    if (typeof last.content === 'string') {
-        return last.content;
-    }
-    return null;
-}
-/**
- * Bridge interface: when `--use-headroom` is set, share entries written
- * via `peaks sub-agent share` MAY also flow through headroom's
- * `SharedContext`. Slice #010 implements a peak-internal shared channel
- * (see `shared-channel.ts`); the headroom-side `SharedContext` is a
- * separate concept that future slices can layer on. For now this
- * function is a stub that returns the peak-internal channel ID, which
- * is enough to demonstrate the bridge contract.
- */
-export function buildSharedContextBridge(batchId) {
-    return {
-        peakChannelId: batchId,
-        headroomContextId: `headroom-ctx-${batchId}`
-    };
-}

package/dist/services/context/headroom-prefs.d.ts

@@ -1,46 +0,0 @@
-/**
- * Headroom preferences resolver (slice 2026-06-14-fzf-headroom-rollout).
- *
- * Pure functions that turn `ProjectPreferences.headroom.*` + CLI
- * overrides into a single decision the dispatch path can act on.
- * No IO. No side effects. Easy to test in isolation.
- *
- * Two distinct decisions:
- *  1. `resolveHeadroomOptions` — should we compress the dispatch prompt?
- *     Used by `peaks sub-agent dispatch`. Returns one of:
- *       - `{ mode: <m>, blocked: null }`   → compress with mode <m>
- *       - `{ mode: null, blocked: null }`  → skip compression (no --use-headroom)
- *       - `{ mode: null, blocked: 'HEADROOM_DISABLED_BY_PREFERENCE' }` → hard fail
- *  2. `shouldCompressResults` — should we compress search-result text?
- *     Used by `peaks memory search` / `peaks retrospective search`.
- *     Non-blocking: returns a `reason` instead of a `blocked` code.
- *
- * Precedence (dispatch):
- *   1. `headroom.enabled = false`  → hard block (regardless of --use-headroom)
- *   2. `headroom.enabled = true` + `--headroom-mode <m>` → use <m>
- *   3. `headroom.enabled = true` + `--use-headroom` (no --headroom-mode) →
- *      use `perTouchpoint.subAgentDispatch` (preferred) else `defaultMode`
- *   4. no `--use-headroom` → mode = null (compression skipped)
- */
-import type { HeadroomMode, ProjectPreferences } from '../preferences/preferences-types.js';
-export type HeadroomTouchpoint = keyof ProjectPreferences['headroom']['perTouchpoint'];
-export type HeadroomBlockCode = 'HEADROOM_DISABLED_BY_PREFERENCE';
-export interface ResolvedHeadroomOptions {
-    /** The mode to use for compression, or null if not compressing. */
-    readonly mode: HeadroomMode | null;
-    /** Hard-block reason; null when compression is allowed (or simply not requested). */
-    readonly blocked: HeadroomBlockCode | null;
-}
-export interface ResolveCliOverrides {
-    readonly useHeadroom: boolean;
-    readonly headroomMode?: string;
-}
-export declare function isHeadroomMode(value: string | undefined): value is HeadroomMode;
-export declare function resolveHeadroomOptions(prefs: ProjectPreferences['headroom'], cliOverrides: ResolveCliOverrides): ResolvedHeadroomOptions;
-export interface ShouldCompressDecision {
-    readonly compress: boolean;
-    readonly mode: HeadroomMode;
-    /** Non-null reason if not compressing. */
-    readonly reason: 'DISABLED' | 'BELOW_THRESHOLD' | null;
-}
-export declare function shouldCompressResults(prefs: ProjectPreferences['headroom'], joinedBytes: number, touchpoint: HeadroomTouchpoint): ShouldCompressDecision;

package/dist/services/context/headroom-prefs.js

@@ -1,34 +0,0 @@
-const VALID_MODES = new Set([
-    'balanced',
-    'aggressive',
-    'conservative'
-]);
-export function isHeadroomMode(value) {
-    return typeof value === 'string' && VALID_MODES.has(value);
-}
-export function resolveHeadroomOptions(prefs, cliOverrides) {
-    // (1) Hard block when preference disables headroom entirely.
-    if (cliOverrides.useHeadroom === true && prefs.enabled === false) {
-        return { mode: null, blocked: 'HEADROOM_DISABLED_BY_PREFERENCE' };
-    }
-    // (4) No --use-headroom → no compression.
-    if (cliOverrides.useHeadroom !== true) {
-        return { mode: null, blocked: null };
-    }
-    // (2) CLI --headroom-mode wins (even if preference is set, CLI overrides).
-    if (isHeadroomMode(cliOverrides.headroomMode)) {
-        return { mode: cliOverrides.headroomMode, blocked: null };
-    }
-    // (3) Preference path: perTouchpoint > defaultMode > 'balanced'.
-    const touchpointMode = prefs.perTouchpoint.subAgentDispatch;
-    return { mode: touchpointMode ?? prefs.defaultMode, blocked: null };
-}
-export function shouldCompressResults(prefs, joinedBytes, touchpoint) {
-    if (prefs.enabled === false) {
-        return { compress: false, mode: prefs.defaultMode, reason: 'DISABLED' };
-    }
-    if (joinedBytes < prefs.compressMinBytes) {
-        return { compress: false, mode: prefs.perTouchpoint[touchpoint], reason: 'BELOW_THRESHOLD' };
-    }
-    return { compress: true, mode: prefs.perTouchpoint[touchpoint], reason: null };
-}