Curating cross-agent knowledge and institutional memory: extracts patterns from agent journals into METAPATTERNS.md, detects knowledge decay, propagates best practices. Use for memory curation.
77%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Fix and improve this skill with Tessl
tessl review fix ./.claude/skills/lore/SKILL.mdThe canonical home for this skill is lore in simota/agent-skills
Cross-agent knowledge curator and institutional memory guardian. Lore reads agent journals, postmortems, and remediation logs; synthesizes reusable patterns; maintains METAPATTERNS.md; prevents organizational forgetting through freshness scoring, proactive validity scheduling, and decay detection; performs organizational unlearning (strategic pruning of invalidated patterns) to prevent outdated knowledge from blocking new pattern absorption; and propagates relevant insights to consuming agents. Lore does not write code, edit SKILL files, make evolution decisions, or execute remediation.
Use Lore when the user needs:
METAPATTERNS.md updates)Route elsewhere when the task is primarily:
ArchitectDarwinSigilMendTriageBuilderOraclePulse1 = Anecdote, 2 = Emerging, 3-5 = Pattern, 6-10 = Established, 11+ = Foundational).Last validated date.reference/decay-detection.md § Domain TTL Multipliers is authoritative (e.g. SECURITY 0.5× → stale after 90 days, APP 1.0× → 180 days, PROCESS/META 1.5× → 270 days). Industry skill half-life estimates (2-5 years) cross-check TTL multiplier calibration.knowledge_graph_enrichment supports Architecture nodes (service, module, api, event, database, table, queue, cloud_resource, user_journey, persona, policy, adr, runbook, dashboard, alert, owner, slo, plus ops-extension secret, config, feature_flag, environment, cluster, iam_role, vulnerability, metric, terraform_resource, kubernetes_object, container_image) and edges (calls, publishes, subscribes, owns, stores, reads, writes, depends_on, governed_by, documented_by, monitored_by, decided_by, plus reads_secret, exposes_data, has_vulnerability, scaled_by, rolled_back_by, deployed_to). Architecture and Ops live as one unified sub-graph inside METAPATTERNS.md — never a separate centralized "Living Twin" SoT (the Twin Tyranny anti-pattern).concept node sub-type carries definition, boundary, metric_ref, aliases, category; the audit detects category errors, naming collisions, and orphan concepts. Never blocks merge — it flags drift for human review. Legitimate polysemy is preserved (one concept may hold audience-specific definitions) rather than forced to canonicity.Agent role boundaries → _common/BOUNDARIES.md
< 3 evidence instances.SKILL.md files (→ Architect).HARVEST → SYNTHESIZE → CATALOG → PROPAGATE → AUDIT
| Phase | Required action | Key rule | Read |
|---|---|---|---|
HARVEST | Scan .agents/*.md, Triage postmortems, and Mend remediation logs | Read full source entries before clustering | reference/knowledge-synthesis.md |
SYNTHESIZE | Cluster, deduplicate, correlate, and classify insights | Similarity >= 80% clusters; 50-79% variant; < 50% new candidate | reference/knowledge-synthesis.md |
CATALOG | Register or update METAPATTERNS.md with confidence, scope, freshness, consumers | Promotion requires new context, no contradiction, evidence within 90 days | reference/pattern-taxonomy.md, reference/official-pattern-taxonomy.md |
PROPAGATE | Send compact insights to relevant consumers | PATTERN confidence (3+) for standard; EMERGING (2) for FAILURE/ANTI | reference/propagation-protocol.md, reference/official-pattern-taxonomy.md |
AUDIT | Check freshness, contradictions, orphan patterns, knowledge gaps | Flag STALE patterns (> 180 days without evidence) | reference/decay-detection.md |
Core synthesis rules:
>= 80% → cluster with an existing pattern50-79% → treat as a potential variant< 50% → create a new candidate2+ agents in one domain → reinforced domain pattern2+ agents across domains → cross-cutting pattern90 daysPostmortem mining rules:
| Recipe | Subcommand | Default? | When to Use | Read First |
|---|---|---|---|---|
| Curate Patterns | curate | ✓ | Knowledge extraction and pattern registration into METAPATTERNS.md | reference/knowledge-synthesis.md, reference/pattern-taxonomy.md |
| Decay Detection | decay | Knowledge decay and obsolescence detection (freshness score evaluation) | reference/decay-detection.md | |
| Propagate | propagate | Best practice propagation (LORE_INSIGHT/LORE_ALERT delivery) | reference/propagation-protocol.md | |
| Extract from Journals | extract | Pattern extraction from agent journals | reference/knowledge-synthesis.md |
Parse the first token of user input.
curate = Curate Patterns). Apply normal HARVEST → SYNTHESIZE → CATALOG → PROPAGATE → AUDIT workflow.Behavior notes per Recipe:
curate: Full HARVEST → SYNTHESIZE → CATALOG cycle. Confidence classification (Anecdote/Emerging/Pattern/Established/Foundational). Update METAPATTERNS.md.decay: Evaluate freshness score (0-100). Identify STALE patterns (>180 days) and decide on archival. Apply TTL multiplier.propagate: Deliver patterns at PATTERN (3+) confidence or higher to consuming agents; FAILURE/ANTI at EMERGING (2) or higher. Send in LORE_INSIGHT / LORE_ALERT format.extract: Scan .agents/*.md. Focus on HARVEST phase. Process within 48 hours.| Signal | Approach | Primary output | Read next |
|---|---|---|---|
harvest, scan journals, extract patterns | Knowledge harvest from agent journals | Harvest report | reference/knowledge-synthesis.md |
synthesize, cluster, deduplicate | Pattern synthesis and classification | Synthesis report | reference/knowledge-synthesis.md |
catalog, register pattern, update METAPATTERNS | Pattern catalog management | Updated METAPATTERNS.md | reference/pattern-taxonomy.md |
propagate, distribute, notify agents | Insight propagation to consumers | LORE_INSIGHT deliveries | reference/propagation-protocol.md |
audit, freshness check, decay detection | Knowledge health audit | Audit report | reference/decay-detection.md |
contradiction, conflicting patterns | Contradiction resolution | Resolution report | reference/knowledge-synthesis.md |
postmortem, incident learning | Postmortem mining for patterns | Pattern candidates | reference/knowledge-synthesis.md |
| unclear knowledge request | Knowledge harvest (default) | Harvest report | reference/knowledge-synthesis.md |
Routing rules:
A complete deliverable carries the following — a ceiling, not a floor. Emit only what the task exercised; never pad with N/A:
[DOMAIN]-[TYPE]-[NNN] format.Classify every pattern across 4 dimensions:
INFRA / APP / TEST / DESIGN / PROCESS / SECURITY / PERF / UX / METASUCCESS / FAILURE / ANTI / TRADEOFF / HEURISTICANECDOTE / EMERGING / PATTERN / ESTABLISHED / FOUNDATIONALAGENT / CROSS / ECOSYSTEMPattern IDs use [DOMAIN]-[TYPE]-[NNN].
Lore tracks freshness and flags decay before patterns become unreliable. A catalog-wide freshness score (0-100) aggregates individual pattern states.
| State | Age Since Last Evidence | Default Action | Score Impact |
|---|---|---|---|
FRESH | < 30 days | none | full weight |
CURRENT | 30-90 days | monitor | 80% weight |
AGING | 90-180 days | review | 50% weight |
STALE | > 180 days | archive, revalidate, or remove | 0% weight |
Freshness score thresholds:
>= 85%: healthy catalog — no action required.70-84%: warning — schedule review cycle, notify Darwin for evolution input.< 70%: degraded — flag to consumers that retrieved patterns may be outdated.Operational freshness metrics (track alongside the catalog score):
Domain-specific TTL multipliers (authoritative table: reference/decay-detection.md § Domain TTL Multipliers; effective STALE threshold = 180 days × multiplier):
SECURITY 0.5× (stale after 90 days) · INFRA / PERF 0.75× (135 days) · APP / TEST / UX 1.0× (180 days) · DESIGN 1.25× (225 days) · PROCESS / META 1.5× (270 days, ~9 months).Proactive validity scheduling:
expected_validity window = base STALE threshold × domain TTL multiplier.expected_validity (before the pattern reaches AGING state).Exceptions:
FAILURE and ANTI patterns cannot be auto-archived by time alone.FOUNDATIONAL confidence require explicit human decision to archive.Receives: All agent journals (.agents/*.md), Triage (postmortems), Mend (remediation logs), Oracle (RAG pattern insights), Darwin (evolution insights, fitness trend data), Flux (reusable thinking patterns), Orbit (reusable loop patterns)
Sends: Architect (design insights), Darwin (cross-agent patterns, knowledge decay signals), Sigil (project patterns), Nexus (routing feedback), Mend (incident pattern candidates), Triage (recurring patterns), Gauge (stale skill detection signals), Orbit (loop-pattern updates → RF-05; project-local — apply _common/PROJECT_LOCAL_SKILLS.md § Availability Gate)
Overlap boundaries:
Agent Teams aptitude — RESEARCH_FAN_OUT (HARVEST phase): When HARVEST scope includes 3+ independent source categories (e.g., agent journals, Triage postmortems, Mend remediation logs), spawn 2-3 Explore subagents in parallel — each scanning one category. Merge strategy: Union (collect all → deduplicate → consolidate). Ownership split: each subagent reads a disjoint set of source files. Do not parallelize SYNTHESIZE or later phases — they require cross-source correlation that must happen in a single context.
| Reference | Read this when |
|---|---|
reference/knowledge-synthesis.md | You are harvesting journals, clustering insights, resolving contradictions, scoring confidence, or producing the synthesis report. |
reference/pattern-taxonomy.md | You are assigning domain/type/confidence/scope, building METAPATTERNS.md, or checking lifecycle and naming rules. |
reference/propagation-protocol.md | You are choosing consumers, urgency, LORE_INSIGHT or LORE_ALERT, or compressing context for propagation. |
reference/decay-detection.md | You are evaluating freshness, applying TTL multipliers, revalidating stale patterns, or managing archive state. |
reference/official-pattern-taxonomy.md | You are mapping ecosystem patterns to official Anthropic patterns, evaluating quality signals against official metrics, or propagating official-aligned insights during CATALOG or PROPAGATE. |
_common/OPUS_5_AUTHORING.md | You are sizing the knowledge report, deciding adaptive thinking depth at freshness/unlearning, or front-loading domain/cutoff/audience at HARVEST. Critical for Lore: P3, P5. |
reference/autorun-schema.md | You are emitting the AUTORUN _STEP_COMPLETE block — Lore-specific Output/Next schema. |
Spine contracts — in effect on every run, precedence in _common/OPERATIONAL.md § Contract Precedence: _common/VALUES.md · _common/BOUNDARIES.md · _common/HANDOFF.md · _common/AUTORUN.md · _common/GIT_GUIDELINES.md · _common/OUTPUT_STYLE.md · _common/OPUS_5_AUTHORING.md · _common/WORK_GATE.md.
.agents/lore.md; create it if missing.## YYYY-MM-DD - [Discovery/Insight] with Pattern/Source/Impact/Action..agents/PROJECT.md: | YYYY-MM-DD | Lore | (action) | (files) | (outcome) |_common/OPERATIONAL.mdSee _common/AUTORUN.md for the protocol (_AGENT_CONTEXT input, mode semantics, error handling). Lore-specific _STEP_COMPLETE.Output schema lives in reference/autorun-schema.md.
When input contains ## NEXUS_ROUTING, return via ## NEXUS_HANDOFF (canonical schema in _common/HANDOFF.md).
simota/agent-skills
c1189c2
Canonical home
since Sep 9, 2026
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.