Дмитрий
15bf46a1c0
Dedicated brainstorm на N1 (S5 controller-writable signal) — найден real out-of-band channel без harness-зависимости через subagent gate-process side-channel. V3.4 architectural shift: BLOCKED-сигнал перенесён из controller-written text (v3.2 regex, v3.3 weak heuristic) в side-channel file `subagent-block-<tool-use-id>.json`: - Path derived из tool_use_id (Anthropic-assigned, controller не контролирует generation) - File в `~/.claude/runtime/*` protected paths (Edit/Write/Delete blocked) - Пишет subagent gate-процесс (отдельный node), не controller text response - Parent gate читает при PreToolUse следующего tool-use после Task return Anti-forgery 3 защитных слоя + 6-attack adversarial pre-check (все blocked). 7 spec edits: - §3.4 переписан: out-of-band block-file channel + anti-forgery + failure modes + adversarial pre-check - §3.1 +`subagent-block-<tool-use-id>.json` explicit-mention - §3.2 расширен: subagent gate ТАКЖЕ пишет block-file (side-channel parallel к inheritance read) - §3.2.0 +Smoke 3: user-run probe для verification block-file write механизма - §8 budget +2h (subagent write 0.5h + parent read 0.5h + lockfile 0.3h + spec 0.5h + integration 0.1h) - §9 — S5 убран из acceptable residual list (теперь CLOSED, не как S8) - §10.2 +schema для subagent-block-<tool-use-id>.json Implementation budget: 19.5-29h (v3.3) → 21.5-31h (v3.4). Главный урок v3.4: controller-writable signals fundamentally fake-able, НО можно вырваться из класса через side-channel write — separate process + harness-derived path + protected file. Это НЕ harness-dependent (работает с любым Claude Code где env-vars пробрасываются + subagent gate стартует с теми же хуками). Brainstorm methodology: superpowers:brainstorming skill — 1 clarifying вопрос (closure level выбран б — out-of-band runtime-файл) → 3 защитных слоя дизайна + 6-attack pre-check → user approval → 7 spec edits. Verify-sentinel: 1179/1179 vitest tools-only GREEN. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Observer infrastructure
Passive evidence-loop for the Лидерра «brain» per ADR-011.
Files
episodes-YYYY-MM.jsonl— append-only JSONL, one line per Stop-event. Schema v2 (schema_version: 2): the 5 mandatory fields +decision_provenance(who chose the node),environment(economy_level / model / post_compaction / session_turn / parallel_session),task_size,task_ref,prompt_signal, and anoutcomethat isunknownat write time (refined by/brain-retro). On an internal hook failure a minimalobserver_errormarker line is written instead of a silent skip. Written bytools/observer-stop-hook.mjsviatools/observer-transcript-parser.mjs.notes/YYYY-MM-DD-<slug>.md— optional MD notes for sessions with qualitative history.STATUS.md— auto-generated dashboard. Regenerated per-commit bytools/status-md-generator.mjs..read-counter.json— C3 observer-of-observer counter. Updated on Read of observer files.dashboard.html+dashboard.js+dashboard-core.js— Brain Dashboard: visualises the episode log over the automation-graph topology (4 views — Карта / Разбор / Лента / Агрегат). Runnpm run brain:dashboard, open the printed localhost URL.dashboard-core.jsis pure logic, unit-tested intools/brain-dashboard-core.test.mjs.
Lifecycle
- Write: every Stop-event appends one JSONL line, parsed from the session transcript (Stop-hook).
- Aggregate:
/brain-retroskill reads JSONL each sprint, proposes regulatory candidates. - Surface:
STATUS.mdshows controllers + monthly stats. - Self-prune: C3 warns if 54 weeks pass without any read of observer files.
Routing-tag discipline
When the user dictates a specific method/node (e.g. «запусти discovery-interview»), Claude must emit one line in its response:
<!-- routing: provenance=user_directed_method node=<chosen> counterfactual=<node Claude would have chosen autonomously> -->
The Stop-hook routing-gate (tools/observer-routing-detector.mjs + routingGateDecision) detects a dictated method; if the tag is missing it returns decision: block, so the turn cannot end without the tag. The gate fires at most once per turn (stop_hook_active guard). This makes decision_provenance reliable — factor analysis can separate a router error from a user-dictated one.
Privacy
PII filter (phone numbers, emails, tokens) is applied before every write — see tools/observer-pii-filter.mjs. gitleaks pre-push also scans observer files as part of full-history sweep.
Don't
- Don't edit
episodes-*.jsonlmanually — it's append-only. - Don't write outside
docs/observer/notes/for hand-curated notes. - Don't change
.read-counter.jsonmanually — it's maintained by hooks.
HK1 pre-check (Pravila ADR-010) — verified 2026-05-19
Before registering tools/observer-stop-hook.mjs on Stop event (Task B5), verified collision against 6-component economy/skill-discipline architecture:
- User-level
~/.claude/settings.jsonalready has Stop hook: agent-type Sonnet-4.6 economy compliance verifier (analyzes transcript for claim-without-evidence violations). - Project-level
.claude/settings.json— Stop slot empty.
Result: no overwrite. observer-stop-hook will be added as command-type entry in project-level Stop array. Project + user scopes are independent slots in Claude Code 2.x — both run on the same Stop event without conflict. The agent verifier (user scope) and the JSONL appender (project scope) have non-overlapping responsibilities.