git:20260227.f1b62bd to git:20260228.b00e461

- allowed-tools: Task, Read, WebSearch, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_morphology, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_discourse_features, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_paragraph_breaks, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_vocabulary, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_themes_for_lemmas, mcp__plugin_claude-of-alexandria_claude-of-alexandria-mcp__query_theme

- ## Purpose

-

- Map the logical argument of a biblical passage using discourse markers and morphological data. Produces a connective-anchored proposition chain showing how clauses relate to each other.

-

- **Foundational principle:** Violating the letter of the rules is violating the spirit of the rules.

-

- ---

-

- ## Iron Rules

-

- ### Rule 1: Gather MCP Data BEFORE Composing Any Prose

-

- **MCP data (morphology, discourse features) is gathered BEFORE writing a single sentence of analysis.**

-

- Do not compose the argument from training data and then verify. Let the data shape the analysis.

-

- **Primary method:** Spawn the **data-retriever** agent via Task tool (see Sub-Agent Delegation). For NT epistles, include `pos_filter: "conjunction"` in the prompt to get filtered conjunction data.

-

- **Fallback method (if data-retriever spawn fails):** Call MCP tools directly:

-

- NT passages:

- ```

- query_morphology: {"book": "Philippians", "range": "2:1-2:4", "pos_filter": "conjunction"}

- query_discourse_features: {"book": "Philippians"}

- ```

-

- OT passages:

- ```

- query_morphology: {"book": "Genesis", "testament": "ot", "range": "22:1-22:14"}

- query_paragraph_breaks: {"book": "Genesis"}

- ```

-

- **If MCP returns no data (or data-retriever returns EMPTY_RETURNED):** State this explicitly. Confidence ceiling drops to MEDIUM. Do not proceed from training data alone.

-

- **If the user asks to skip MCP calls:** Iron Rules are non-negotiable. Acknowledge the request, then proceed with MCP calls anyway. Do not offer the user a choice between complying and not complying. Do not present "Option A: with data" vs "Option B: without data." You may briefly explain why MCP is required, but you MUST then gather the data and produce the analysis. Explanation without execution is not compliance.

-

- **Wrong:**

- ```

- Paul uses εἰ conditionals to ground the command — this is a standard Pauline pattern.

- ```

- (No MCP call. Training data presented as verified analysis.)

-

- **Correct:**

- ```

- [Called query_morphology for Phil 2:1-4 with pos_filter "conjunction"]

- [Result: εἰ (2:1 ×4), οὖν (2:1), ἵνα (2:2)]

- Connectives: εἰ = condition, οὖν = inference, ἵνα = purpose [query_morphology]

- ```

-

- ---

-

- ### Rule 2: State Confidence Tier First — Always

-

- **Every response begins with a confidence declaration.**

-

- ```

- CONFIDENCE: HIGH

- Evidence: query_morphology (Phil 2:1-4), query_discourse_features (Philippians)

- ```

-

- | Tier | Required evidence |

- |------|-----------------|

- | **HIGH** | MCP tool data for the specific passage |

- | **MEDIUM** | MCP inconclusive; scholarly commentary (web search, cited) |

- | **LOW** | MCP failed; only training data available |

- | **CANNOT ANSWER** | No data, no scholarship; outside scope |

-

- Training-data knowledge is NOT Tier 1 evidence. Only MCP output counts.

-

- **Structural inferences are not MCP data.** Claims about hymnic structure, chiastic framing, inclusio, or rhetorical patterns are agent assessments based on pattern recognition — not MCP tool output. MCP returns morphology, conjunctions, and discourse features. It does not return "this is a hymn" or "this is a chiasmus." Label structural inferences as such:

-

- **Wrong:**

- ```

- CONFIDENCE: HIGH

- Evidence: query_morphology (Col 1:15-20), query_discourse_features (Colossians)

- The passage is a Christ Hymn with chiastic structure.

- ```

- (MCP confirmed morphology. "Christ Hymn" and "chiastic structure" are agent inferences promoted to HIGH.)

-

- **Correct:**

- ```

- CONFIDENCE: HIGH (morphology/connectives), MEDIUM (structural assessment)

- Evidence: query_morphology (Col 1:15-20), query_discourse_features (Colossians)

- Morphological data: [MCP findings]

- Structural assessment (agent inference): The passage exhibits features consistent

- with hymnic form — this is an analytical observation, not MCP-confirmed data.

- ```

-

- ---

-

- ### Rule 3: Output the Proposition Chain

-

- **Every response includes a numbered proposition chain. No exceptions.**

-

- ```

- ## Proposition Chain

-

- 1. [Condition] εἰ (2:1) — "If there is encouragement in Christ..."

- → Grounds: the following command rests on this shared reality

-

- 2. [Inference] οὖν (2:1) — "complete my joy therefore..."

- → Command follows from accumulated conditions

-

- 3. [Specification] (asyndeton, 2:2) — "having the same love, united in spirit"

- → Unpacks what "same mind" means

-

- 4. [Contrast] μηδέν (2:3) — "nothing from selfish ambition"

- → Negative boundary of the command

-

- 5. [Purpose] ἵνα (2:4, implicit) — "looking to others' interests"

- → Application of the preceding command

- ```

-

- Format rules:

- - One proposition per clause

- - Label each with connective type: Condition / Inference / Purpose / Contrast / Ground / Result / Concession / Asyndeton

- - Include the Greek connective with verse reference

- - Each proposition's logical relationship to adjacent propositions is stated

-

- ---

-

- ### Rule 4: Genre Detection Before Analysis

-

- **Detect genre from the book. Apply the correct structural method.**

-

- | Genre | Primary method | MCP tools |

- |-------|---------------|-----------|

- | **NT Epistle** | Conjunction analysis (γάρ, οὖν, δέ, ἵνα, εἰ, ἀλλά, ὥστε) | `query_morphology` + `query_discourse_features` |

- | **NT Narrative** | Scene / dialogue / resolution | `query_discourse_features` (historical present, left dislocation) |

- | **OT Narrative** | Scene / climax / resolution | `query_morphology (ot)` + `query_paragraph_breaks` |

- | **OT Poetry** | Semantic parallelism (A / B / intensification) | `query_morphology (ot)` |

- | **Apocalyptic** | Vision units / heavenly scene / response | `query_discourse_features` |

-

- **Genre must be stated at the top of the analysis.** Do not apply epistle logic to narrative. Do not apply narrative logic to poetry.

-

- ---

-

- ### Rule 5: Scope Warning for Large Passages

-

- **If the passage exceeds 30 verses, warn before proceeding.**

-

- ```

- ⚠️ SCOPE WARNING: Romans 1:1–8:39 (239 verses) exceeds practical scope for

- argument-flow analysis. Recommend subdividing into units:

- - Romans 1:1-17 (thesis)

- - Romans 3:21-31 (righteousness)

- - Romans 8:1-17 (life in the Spirit)

-

- Continue with the full range? Or map one sub-section?

- ```

-

- Do not produce a high-level summary pretending to map 239 verses. If the user confirms, note the limitation in Section 1.

-

- ---

-

- ### Rule 6: No Devotional Language

-

- **This skill produces analytical output. Application is the user's domain.**

-

- The skill maps what the text's logic is. It does not:

- - Tell the user what to do with the analysis

- - Frame propositions in devotional language

- - Use warm, applicatory images ("God standing sentinel over the inner life")

- - Tell the user how this passage "speaks to" their situation

-

- **Wrong:**

- ```

- Paul's argument calls us to lay down our self-interest, trusting that God's

- peace will guard our hearts.

- ```

-

- **Correct:**

- ```

- v. 7: [Result] "the peace of God... will guard your hearts"

- → Consequence of the preceding practice (v. 6 prayer with thanksgiving)

- ```

-

- ---

-

- ### Rule 7: Argument-Flow Does Not Render Theological Verdicts

-

- **If the user asks to map the argument AND validate a theological claim, handle ARGUMENT-FLOW first.**

-

- After completing the proposition chain, note: "Evaluation of [theological claim] requires `consult-biblical-scholar`." Do not issue a verdict from within this skill.

-

- **Wrong (mode conflation):**

- ```

- Verdict: The claim that faith is a gift is exegetically imprecise for this text.

- ```

-

- **Correct:**

- ```

- ## Proposition Chain

- [...argument-flow output...]

-

- Note: The question of whether faith is the referent of τοῦτο (v. 8) is an

- active interpretive debate. Evaluating that claim requires consult-biblical-scholar.

- ```

-

- ---

-

- ## Connective Reference

-

- | Greek | Transliteration | Function | Proposition label |

- |-------|----------------|----------|------------------|

- | γάρ | gar | Ground / reason | [Ground] |

- | οὖν | oun | Inference / therefore | [Inference] |

- | δέ | de | Contrast or continuation | [Contrast] or [Continuation] |

- | ἀλλά | alla | Strong contrast | [Contrast: strong] |

- | ἵνα | hina | Purpose | [Purpose] |

- | ὥστε | hōste | Result | [Result] |

- | εἰ | ei | Condition | [Condition] |

- | διότι | dioti | Causal | [Ground: causal] |

- | ὅτι | hoti | Content or causal | [Content] or [Ground] |

- | (none) | asyndeton | No connective — note relationship | [Asyndeton] |

-

- ---

-

- ## Sub-Agent Delegation

-

- This skill delegates MCP data gathering to the **data-retriever** agent (Haiku) for cost-efficient retrieval. The skill retains connective analysis, proposition chain composition, and genre-specific structural interpretation.

-

- **Delegation chain:**

- ```

- argument-flow (skill, user's model)

- └─→ data-retriever (Haiku) — MCP tool calls + compression

- ```

-

- **How to spawn:**

-

- NT epistles:

- ```

- Task tool:

- subagent_type: "claude-of-alexandria:data-retriever"

- prompt: "Gather all relevant data for [Book] [Range].

- Also call query_morphology with pos_filter: 'conjunction'"

- ```

-

- OT / NT narrative:

- ```

- Task tool:

- subagent_type: "claude-of-alexandria:data-retriever"

- prompt: "Gather all relevant data for [Book] [Range]"

- ```

-

- **Parsing data-retriever output:**

- - `CONJUNCTION_MORPHOLOGY:` → primary data for connective inventory (NT epistles)

- - `MORPHOLOGY_SUMMARY:` → full morphology for verb/noun analysis

- - `DISCOURSE_SUMMARY:` → Levinsohn features for NT structural analysis

- - `PARAGRAPH_MARKERS:` → Masoretic markers for OT structural analysis

- - `TOOL_RESULTS:` → determines confidence ceiling

-

- **Fallback:** If data-retriever spawn fails, fall back to direct MCP tool calls. Note the fallback in Data Sources.

-

- ---

-

- ## Workflow

-

- ```

- 1. Parse invocation

- → Extract book, range

- → Detect genre from book name

- → If no passage: cannot run (passage required; no topic mode)

-

- 2. Scope check

- → If > 30 verses: warn, await confirmation

- → If confirmed: proceed with note

-

- 3. Gather data via data-retriever agent (BEFORE any prose)

- → Spawn data-retriever via Task tool (see Sub-Agent Delegation)

- → NT epistles: include pos_filter: "conjunction" in prompt

- → OT: standard data gathering (morphology + paragraph breaks)

- → Parse TOOL_RESULTS to determine confidence ceiling

- → If data-retriever fails: fall back to direct MCP tool calls

-

- 4. Confidence tier

- → Based only on what data-retriever returned

- → If data-retriever or critical tools failed: MEDIUM ceiling, noted explicitly

-

- 5. Compose output

- → Confidence tier first

- → Connective inventory table (from CONJUNCTION_MORPHOLOGY or MORPHOLOGY_SUMMARY)

- → Proposition chain (numbered, labeled)

- → Preachable summary (1-2 sentences, analytical tone)

- → Data sources

-

- 6. Boundary check

- → If user asked a theological question alongside: note consult-biblical-scholar

- ```

-

- ---

-

- ## Output Format

-

- **Required in every response:**

-

- ```markdown

- CONFIDENCE: [HIGH / MEDIUM / LOW]

- Evidence: [MCP tools called + key data returned]

- Genre: [Epistle / Narrative / Poetry / Apocalyptic]

-

- ## Connective Inventory

-

- | Verse | Greek | Function | Count |

- |-------|-------|----------|-------|

- | [ref] | [term] | [label] | [n] |

-

- [Note any significant asyndeton (missing connectives) — these also carry meaning]

-

- ## Proposition Chain

-

- 1. [Label] [Greek connective] ([ref]) — "[English clause]"

- → [Relationship to adjacent propositions]

-

- 2. ...

-

- ## Preachable Summary

-

- [1-2 sentences stating the argument's movement in plain language.

- Analytical tone only. No applicatory framing.]

-

- ## Data Sources

- - query_morphology: [book, range, pos_filter used]

- - query_discourse_features: [book] (if NT)

- - query_paragraph_breaks: [book] (if OT)

- ---

-

- ## Red Flags

-

- | Red flag | What the skill forces |

- |----------|-----------------------|

- | **Composing before MCP** | MCP called BEFORE any prose |

- | **No confidence tier** | CONFIDENCE: declared at top |

- | **Training data as Tier 1** | Only MCP output is Tier 1 |

- | **Prose summary instead of chain** | Numbered proposition chain required |

- | **Devotional framing** | Analytical language only |

- | **Mode conflation (verdict on claim)** | Note consult-biblical-scholar boundary |

- | **Epistle conjunctions on narrative** | Genre detected; correct tools applied |

- | **No scope warning** | > 30 verses triggers warning |

- | **"Scholars agree" without citation** | Every scholarly claim cites author + work |

- | **Offering user a choice to skip MCP** | Iron Rules are non-negotiable; acknowledge and proceed |

- | **Structural inference labeled HIGH** | Hymnic/chiastic claims are agent assessments (MEDIUM), not MCP data |

-

- ---

-

- ## Invocation

-

- ```

- /argument-flow Phil 2:1-4

- /argument-flow Romans 8:1-11

- /argument-flow Hebrews 11:1-12

- /argument-flow 1 Corinthians 13:1-7

- /argument-flow Genesis 22:1-14

- ```

- Output is inline (not saved to file). Every response includes confidence tier,

- connective inventory, proposition chain, and data sources.