CLI: surrogate-safe truncation for hook tool-arg bounds + task-runtime detail (remaining raw slices) #870

Self-discovery (undirected maintainer review pass). Left unclaimed for the next natural tick.

sourceType

Self-discovery (repository maintainer review of qwen-code-dev-bot/oh-my-cli).

sourceLinkOrEvidence

• src/hook-contract.ts:506 — boundForHook: return value.length <= maxLen ? value : ${value.slice(0, maxLen)}…; — raw slice(0, maxLen) on arbitrary tool-call argument strings.
• src/hook-contract.ts:499 — the same file's reason-truncator already uses safeCutEnd (text.slice(0, safeCutEnd(text, MAX_REASON - 1))), so boundForHook is a same-file inconsistency; safeCutEnd is already imported (src/hook-contract.ts:31).
• src/task-runtime.ts:777 — safeText: return trimmed.slice(0, Math.max(0, max - 1)) + "…"; — raw slice on redacted task-detail text (MAX_DETAIL_CHARS = 200); safeCutEnd is not yet imported there.

Minimal reproduction: boundForHook("a".repeat(99) + "🚀 tail", 100) keeps indices 0-99 via slice(0, 100), retaining the rocket's high surrogate (index 99) while cutting its low partner (index 100) — an orphaned high surrogate before the …. The same shape applies to safeText at the MAX_DETAIL_CHARS - 1 bound.

problemStatement

Two truncation sites that handle arbitrary/user-authored text still cut with a raw String.prototype.slice(0, N): boundForHook (tool-call arguments forwarded to hook stdin) and safeText (task detail). When an astral character straddles the cut boundary, the kept prefix ends in an orphaned surrogate — producing malformed text in the hook payload / task detail. This breaks the surrogate-safe safeCutEnd convention established by #812/#821/#828 and extended through #860/#868.

userValue

Hook payloads and task details containing emoji/astral characters truncate cleanly instead of ending in an unpaired surrogate — consistent with every other truncation path in the CLI.

scope

• src/hook-contract.ts boundForHook: cut via safeCutEnd (already imported; mirror the sibling reason-truncator).
• src/task-runtime.ts safeText: import safeCutEnd and cut via safeCutEnd(trimmed, Math.max(0, max - 1)).
• Keep the … markers and the existing length bounds unchanged.
• Red→green regression tests for both sites.

nonGoals

• Not changing length bounds, markers, redaction, or the hook/tool runtime behavior.
• Command/name display paths with lower astral probability (repo-context.ts redactName/redactCommand, repo-readiness.ts:125, run-bottleneck.ts:59 safeName) are out of scope here; file separately if warranted.
• Byte-bounded output caps (already surrogate-safe via Robustness: make byte-bounded truncations surrogate-safe (headless output + change-review path/name) #828/Robustness: make the hook output-size cap surrogate-safe (spawnHookRunner) #830/Robustness: make the tool-invocation output-size cap surrogate-safe (spawnCommandRunner) #832/Robustness: account subprocess output caps in UTF-8 bytes, not UTF-16 code units (hook/tool/MCP) #834/Robustness: reassemble multi-byte UTF-8 split across chunks in the hook/tool/MCP output caps #836/Robustness: make the provider-invocation output cap UTF-8-byte-accurate and surrogate-safe (capOutput) #848) are untouched.

acceptanceCriteria

• Given a tool-argument string / task detail whose astral character straddles the truncation boundary, the truncated result contains no orphaned surrogate (regex [\ud800-\udbff](?![\udc00-\udfff])|(?<![\ud800-\udbff])[\udc00-\udfff] does not match) and still ends with the existing … marker.
• boundForHook and the hook-contract reason-truncator use the same safeCutEnd mechanism.
• Inputs at or below the bound are returned unchanged.

testPlan

• Regression tests in tests/unit/hook-contract.test.ts and tests/unit/task-runtime.test.ts: surrogate-straddle at each bound, in-range astral passthrough, and long-ASCII truncation.
• Red→green: reverting the two cuts to raw slice makes the new tests fail; restoring safeCutEnd turns them green.
• Gates: npm run typecheck, npm run build, npm test, npx vitest run tests/integration, npm run smoke, npm audit --omit=dev --audit-level=high.

dogfoodPlan

After merge, run npx vitest run tests/unit/hook-contract.test.ts tests/unit/task-runtime.test.ts tests/unit/text-cut.test.ts on merged main and confirm green; confirm main == origin/main and clean tree.

riskAndSecurityNotes

Behavior surface unchanged except surrogate safety at the cut; redaction and hook/tool runtime semantics are untouched. Tests use inert local strings (no secrets). No new attack surface.

duplicateSearchEvidence

• gh search issues "boundForHook" / "task-runtime truncat" / "hook truncation surrogate" -> no results.
• gh search issues "safeText surrogate" -> only Robustness: make byte-bounded truncations surrogate-safe (headless output + change-review path/name) #828 (closed), which covers the headless-output + change-review path/name byte caps — different sites.
• CLI: surrogate-safe tail truncation for task-verify failure output and failure receipts #860/CLI: surrogate-safe truncation for goal objective + workspace-memory text (remaining raw slices deferred by #860) #868 covered task-verify/failure-receipts and session-goal/workspace-memory respectively; neither covered hook-contract or task-runtime.

parentChildRelationship

Standalone leaf subtask; no parent roadmap item required and no children anticipated.

dependencyOrder

No dependencies; dependency-ready. Self-contained change to hook-contract.ts and task-runtime.ts (+ tests) using the existing text-cut.ts helper.