菜单
Advanced semantic vocabulary, holographic contracts, and structural patterns for OCTAVE. REQUIRES octave-literacy. Extends literacy with mythology, archetype annotation, v6 contracts, and anti-pattern rules.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Supabase test harness patterns including local Supabase setup, test user creation via Auth Admin API, RLS testing, migration testing (db_reset), seed sequences, rate limiting, and environment detection. Use when setting up Supabase testing infrastructure, creating test users, troubleshooting Supabase test failures, or implementing RLS validation. Triggers on: supabase test setup, test user creation, supabase local testing, RLS testing, migration testing, supabase test harness, auth test helpers.
Supabase operational knowledge for migrations, RLS optimization, MCP tool benchmarks, and ADR-003 compliance. Use when validating database migrations, optimizing Row-Level Security policies, checking MCP tool performance, or ensuring Supabase operational standards. Triggers on: migration validation, RLS patterns, Supabase benchmarks, ADR-003, database state tracking, schema governance.
LLM-native structured communication format. Teaches OCTAVE syntax rules, canonical forms, and warning prevention for zero-error .oct.md authoring.
Workflow for transforming prose into semantic OCTAVE structures. Covers tier selection, transformation phases, loss accounting, and decision rules. REQUIRES octave-literacy.
Strategic oversight mode for the Holistic Orchestrator. Episodic State Machine pattern with ho-liaison advisory (via pal_clink), run_debate escalation, and oa-router delegation. REPLACES holistic-orchestration when loaded.
Proper delegation patterns for Task() invocations with governance context injection
| name | octave-mastery |
| description | Advanced semantic vocabulary, holographic contracts, and structural patterns for OCTAVE. REQUIRES octave-literacy. Extends literacy with mythology, archetype annotation, v6 contracts, and anti-pattern rules. |
| allowed-tools | ["Read","Write","Edit"] |
| triggers | ["octave architecture","agent design","semantic pantheon","advanced octave","OCTAVE mastery","holographic patterns","archetypes","high-density specifications","system architecture","holographic contracts","archetype annotation"] |
| version | 3.1.0 |
STATUS::["ACTIVE"∧ENUM[ACTIVE,DRAFT,DEPRECATED]→§LIFECYCLE] PRIORITY::["HIGH"∧ENUM[HIGH,MEDIUM,LOW]] LATENCY::["<200ms"∧REGEX[<\d+ms]]
CONSTRAINT_TYPES:
CORE::[
REQ,
OPT,
CONST,
ENUM,
TYPE,
REGEX,
DIR,
APPEND_ONLY
]
EXTENDED::[
RANGE,
MAX_LENGTH,
MIN_LENGTH,
DATE,
ISO8601
]
§4c::CONTRACT_BLOCK_EXAMPLE
// Full META block using CONTRACT for a typed document
EXAMPLE:
```
===MY_DOCUMENT===
META:
TYPE::DECISION
VERSION::"1.0.0"
CONTRACT::HOLOGRAPHIClatest@localSTATUS::["ACTIVE"∧ENUM[ACTIVE,DRAFT,DEPRECATED]]
OWNER::["team-name"∧REQ]
DECISION::["adopt microservices"∧REQ]
===END===
§5::BLOCK_NOTATION_RULE // Hierarchical content MUST use block notation RULE::"Nested structures (maps containing maps) MUST use BLOCK notation: single colon + indented children" NEVER::"Inner value is itself an inline map: KEY::[outer::[inner::val]] — error E_NESTED_INLINE_MAP in strict mode, warning W_NESTED_INLINE_MAP in lenient mode" PRIMARY_CASE::"Agent §2::BEHAVIOR definitions — CONDUCT, PROTOCOL, OUTPUT blocks require block notation" CORRECT:
CONDUCT:
TONE::"Precise"
PROTOCOL:
MUST_ALWAYS::[rule_a, rule_b]
```
WRONG::"CONDUCT::[TONE::"Precise", PROTOCOL::[MUST_ALWAYS::[rule_a]]]"
RECOVERY_ON_E_NESTED_INLINE_MAP::"Fix is BLOCK form — NOT flatten-to-scalars (creates FLAT_PREFIX_SCALARS anti-pattern, §6), NOT collapse-to-blob (creates W_SNAKE_CASE_BLOB). The wrong escapes produce a different error class, not a solution."
§6::ANTI_PATTERNS
// Each has a concrete example of what NOT to do
ISOLATED_LIST::"[auth, payments, users] with no relationships — use DECISION::microservice_extraction[auth⊕payments→independent_services] instead"
FLAT_HIERARCHY::"All keys at top level with no grouping — group related keys under a parent BLOCK"
BURIED_NETWORK::"RELATED_TO::other_service hidden in prose comment — use explicit operator: auth→payments[dependency]"
OPERATOR_SOUP::"RESULT::A+B->C~D all in one expression — break into separate keyed fields"
PROSE_BLEED::"Using natural language sentences as values — every token must carry semantic payload"
INLINE_ARRAY_ROOT::"TOKEN::[KEY::v, KEY2::v2] — inline map used as multi-field token root. Non-canonical for map tokens even when values are atomic; BLOCK form is mandated. Triggers E_NESTED_INLINE_MAP as soon as any child value needs nesting (§5::BLOCK_NOTATION_RULE). Fix: use BLOCK form (TOKEN: + indented children) from the start."
FLAT_PREFIX_SCALARS::"PARENT_CHILD::v, PARENT_CHILD2::v2 — flattened hierarchy via key name prefixes instead of BLOCK nesting. Destroys grouping and LLM attention. Fix: group under a PARENT: block with CHILD::v children."
§7::TIER_NORMALIZATION_AUDIT_CHANNEL
// ADR-0006 SR1-T1 Step 3 (v1.12.0): centralised audit channel for I4 completeness
MODULE::"octave-mcp:src/octave_mcp/core/grammar/tier_normalize.py"
// ^ path is in the octave-mcp repo (upstream OCTAVE implementation), not this repo.
API:
LOG_REPAIR::"log_repair(log, rule_id, before, after, *, safe=True, semantics_changed=False) — single precise entry point appending a RepairTier.NORMALIZATION entry"
ACTIVE::"active(log) context manager — binds log as the ContextVar-scoped sink so pipeline-internal sites (notably emitter identifier-dequoting) can record receipts without threading RepairLog through emit()'s public signature"
RECONCILE::"reconcile_canonical_emission(log, baseline_bytes, canonical_bytes) — reconciler bridge"
RULE_IDS:
PRECISE::"TN_IDENTIFIER_DEQUOTE — was_quoted-driven dequoting at emit_assignment"
BRIDGE::"TN_RECONCILE_CANONICAL — coarse-grained post-emit baseline-vs-canonical receipt"
RECONCILER_BRIDGE_PATTERN:
PURPOSE::"Closes the audit-cardinality gap for diffs that upstream precise loggers do not yet cover (blank-line stripping pending Sprint 3+ trivia population; triple-quote collapse pending new lexer W-code)"
DEDUP_DISCRIMINANT::"Only RepairTier.NORMALIZATION entries on the log suppress the bridge. RepairTier.REPAIR (schema repairs) and RepairTier.FORBIDDEN entries do NOT count — they are orthogonal dimensions"
SELF_DEPRECATION::"When Sprint 3+ trivia and the new triple-quote lexer W-code land, precise loggers will cover their respective diffs, the dedup precondition fails, and the reconciler no-ops without any code change"
INTEGRATION_POINTS:
EMITTER::"emit_assignment: precise log when assignment.was_quoted is True AND emitter chose to emit bare (identifier-shape dequoting)"
WRITE_PY::"octave-mcp:src/octave_mcp/mcp/write.py — tier_normalize.active(tier_normalize_log) wraps each _emit_with_style call; reconcile_canonical_emission runs after final emit; entries drain into result[corrections]"
===END===