| 1 | # Knowledge Engineering Knowledge Set v1 (Operations Rules) |
| 2 | |
| 3 | Set: 20 cards. |
| 4 | Focus: knowledge-engineering operational reliability (retention, section updates, restore protocol, archive discipline). |
| 5 | |
| 6 | --- |
| 7 | |
| 8 | ## Card 01 |
| 9 | - card_id: KC-2026-02-26-159 |
| 10 | - world: knowledge-engineering.operations |
| 11 | - epistemic_basis: `fact` |
| 12 | - confidence: `high` |
| 13 | - signal: full-file overwrite risks accidental memory loss in collaborative edits |
| 14 | - action: prefer section upsert or append for routine updates |
| 15 | - transfer_boundary: full write allowed only with read-merge-write guardrail |
| 16 | |
| 17 | ## Card 02 |
| 18 | - card_id: KC-2026-02-26-160 |
| 19 | - world: knowledge-engineering.operations |
| 20 | - epistemic_basis: `fact` |
| 21 | - confidence: `high` |
| 22 | - signal: section markers enable deterministic targeted updates |
| 23 | - action: keep stable `section_id` contracts for operational blocks |
| 24 | - transfer_boundary: marker drift invalidates automated upsert safety |
| 25 | |
| 26 | ## Card 03 |
| 27 | - card_id: KC-2026-02-26-161 |
| 28 | - world: knowledge-engineering.revisions |
| 29 | - epistemic_basis: `fact` |
| 30 | - confidence: `high` |
| 31 | - signal: revision volume grows rapidly in active multi-agent workflow |
| 32 | - action: enforce retention windows (keep-last/daily/weekly/monthly) |
| 33 | - transfer_boundary: retention must protect referenced audit anchors |
| 34 | |
| 35 | ## Card 04 |
| 36 | - card_id: KC-2026-02-26-162 |
| 37 | - world: knowledge-engineering.revisions |
| 38 | - epistemic_basis: `fact` |
| 39 | - confidence: `high` |
| 40 | - signal: uncontrolled archive growth reduces retrieval usability |
| 41 | - action: trigger compaction by threshold (count/size) with pre-summary |
| 42 | - transfer_boundary: never compact without preserving decision-critical lineage |
| 43 | |
| 44 | ## Card 05 |
| 45 | - card_id: KC-2026-02-26-163 |
| 46 | - world: knowledge-engineering.loading |
| 47 | - epistemic_basis: `fact` |
| 48 | - confidence: `high` |
| 49 | - signal: full archive load at chat start causes noise and slows routing |
| 50 | - action: load hot context first, fetch deep archive on explicit need |
| 51 | - transfer_boundary: deep pull only for missing facts/history |
| 52 | |
| 53 | ## Card 06 |
| 54 | - card_id: KC-2026-02-26-164 |
| 55 | - world: knowledge-engineering.loading |
| 56 | - epistemic_basis: `fact` |
| 57 | - confidence: `high` |
| 58 | - signal: context compression events break continuity in long threads |
| 59 | - action: restore via artifact path (index -> playbook -> latest batch) |
| 60 | - transfer_boundary: do not rely on chat memory as primary recovery source |
| 61 | |
| 62 | ## Card 07 |
| 63 | - card_id: KC-2026-02-26-165 |
| 64 | - world: knowledge-engineering.indexing |
| 65 | - epistemic_basis: `fact` |
| 66 | - confidence: `high` |
| 67 | - signal: new files without index linkage become practically invisible |
| 68 | - action: synchronize README + canonical knowledge-index on each batch |
| 69 | - transfer_boundary: batch is not complete until both indexes are updated |
| 70 | |
| 71 | ## Card 08 |
| 72 | - card_id: KC-2026-02-26-166 |
| 73 | - world: knowledge-engineering.indexing |
| 74 | - epistemic_basis: `inference` |
| 75 | - confidence: `medium` |
| 76 | - signal: long index blocks become hard to scan under pressure |
| 77 | - action: keep status artifact per domain for fast restart |
| 78 | - transfer_boundary: status artifact supplements, not replaces detailed index |
| 79 | |
| 80 | ## Card 09 |
| 81 | - card_id: KC-2026-02-26-167 |
| 82 | - world: knowledge-engineering.classification |
| 83 | - epistemic_basis: `fact` |
| 84 | - confidence: `high` |
| 85 | - signal: mixed-world cards reduce reusability and increase contradiction risk |
| 86 | - action: one card -> one world -> explicit transfer boundary |
| 87 | - transfer_boundary: cross-world fusion requires conflict note |
| 88 | |
| 89 | ## Card 10 |
| 90 | - card_id: KC-2026-02-26-168 |
| 91 | - world: knowledge-engineering.classification |
| 92 | - epistemic_basis: `inference` |
| 93 | - confidence: `medium` |
| 94 | - signal: operational and narrative fragments frequently co-mingle |
| 95 | - action: move narrative history to cold archive, keep hot layer operational |
| 96 | - transfer_boundary: narrative can return only with explicit task need |
| 97 | |
| 98 | ## Card 11 |
| 99 | - card_id: KC-2026-02-26-169 |
| 100 | - world: knowledge-engineering.quality-gates |
| 101 | - epistemic_basis: `fact` |
| 102 | - confidence: `high` |
| 103 | - signal: high-confidence claims appear with weak provenance |
| 104 | - action: block promotion without source and falsification trigger |
| 105 | - transfer_boundary: confidence cannot exceed evidence class |
| 106 | |
| 107 | ## Card 12 |
| 108 | - card_id: KC-2026-02-26-170 |
| 109 | - world: knowledge-engineering.quality-gates |
| 110 | - epistemic_basis: `fact` |
| 111 | - confidence: `high` |
| 112 | - signal: duplicate cards emerge during rapid batching |
| 113 | - action: run periodic dedupe with supersedes/superseded_by fields |
| 114 | - transfer_boundary: dedupe cannot delete unique boundary information |
| 115 | |
| 116 | ## Card 13 |
| 117 | - card_id: KC-2026-02-26-171 |
| 118 | - world: knowledge-engineering.multi-agent |
| 119 | - epistemic_basis: `fact` |
| 120 | - confidence: `high` |
| 121 | - signal: concurrent edits by rotating agents create race-like update risks |
| 122 | - action: favor additive updates + section upserts over monolithic rewrites |
| 123 | - transfer_boundary: monolithic rewrite requires explicit lock/protocol |
| 124 | |
| 125 | ## Card 14 |
| 126 | - card_id: KC-2026-02-26-172 |
| 127 | - world: knowledge-engineering.multi-agent |
| 128 | - epistemic_basis: `inference` |
| 129 | - confidence: `medium` |
| 130 | - signal: trust-driven "continue without stopping" mode increases throughput and drift risk |
| 131 | - action: pair large batches with strict completion checklist |
| 132 | - transfer_boundary: speed mode invalid without post-batch sync checks |
| 133 | |
| 134 | ## Card 15 |
| 135 | - card_id: KC-2026-02-26-173 |
| 136 | - world: knowledge-engineering.restore |
| 137 | - epistemic_basis: `fact` |
| 138 | - confidence: `high` |
| 139 | - signal: environment-level context compression can feel like abrupt reset |
| 140 | - action: maintain compact domain-status artifacts for deterministic restart |
| 141 | - transfer_boundary: status must be updated before declaring domain done |
| 142 | |
| 143 | ## Card 16 |
| 144 | - card_id: KC-2026-02-26-174 |
| 145 | - world: knowledge-engineering.restore |
| 146 | - epistemic_basis: `fact` |
| 147 | - confidence: `high` |
| 148 | - signal: restoration from transcript alone is costly and noisy |
| 149 | - action: prioritize canonical notes and indexed knowledge files as source of truth |
| 150 | - transfer_boundary: transcript mining reserved for unresolved ambiguity |
| 151 | |
| 152 | ## Card 17 |
| 153 | - card_id: KC-2026-02-26-175 |
| 154 | - world: knowledge-engineering.evidence |
| 155 | - epistemic_basis: `fact` |
| 156 | - confidence: `high` |
| 157 | - signal: operational lessons emerge from tool traces, not only prose |
| 158 | - action: treat tool-call history as first-class provenance |
| 159 | - transfer_boundary: untraceable claims remain provisional |
| 160 | |
| 161 | ## Card 18 |
| 162 | - card_id: KC-2026-02-26-176 |
| 163 | - world: knowledge-engineering.evidence |
| 164 | - epistemic_basis: `inference` |
| 165 | - confidence: `medium` |
| 166 | - signal: archive extract quality depends on focused query strategy |
| 167 | - action: use targeted keyword passes and normalize into cards |
| 168 | - transfer_boundary: broad sweep without normalization is not promotable |
| 169 | |
| 170 | ## Card 19 |
| 171 | - card_id: KC-2026-02-26-177 |
| 172 | - world: knowledge-engineering.governance |
| 173 | - epistemic_basis: `fact` |
| 174 | - confidence: `high` |
| 175 | - signal: domain closure without maintenance policy leads to silent decay |
| 176 | - action: add refresh triggers and handoff statement per domain |
| 177 | - transfer_boundary: closure status expires without refresh cadence |
| 178 | |
| 179 | ## Card 20 |
| 180 | - card_id: KC-2026-02-26-178 |
| 181 | - world: knowledge-engineering.governance |
| 182 | - epistemic_basis: `fact` |
| 183 | - confidence: `high` |
| 184 | - signal: cleanup pressure appears after rapid expansion |
| 185 | - action: split "finish domain" and "cleanup pass" into separate phases |
| 186 | - transfer_boundary: no structural cleanup during active evidence ingestion |
| 187 | |
| 188 | --- |
| 189 | |
| 190 | ## Source Footing |
| 191 | |
| 192 | - Archive revision extracts (retention, revision hygiene, layered loading). |
| 193 | - Tool-trace patterns from long operational transcript (append/upsert/restore behavior). |
| 194 | - Operational incidents around context compression and recovery workflow. |
| 195 | |
| 196 | |