| name | douglas-adamiser |
| description | This skill should be used when the user asks to "write like Douglas Adams", "transform text into Adams style", "Hitchhiker's Guide style", "make this funnier", "add absurdist humor", "write sci-fi comedy", "parody this", or mentions Douglas Adams, bathos, anti-climax, or British absurdist humor. Transforms any text into Douglas Adams' distinctive style with bathos, cosmic irony, philosophical tangents, and perfectly timed anti-climaxes. Includes validation scripts and vocabulary checks. |
Douglas Adamiser
Purpose
Transform any text (news articles, academic papers, technical documentation, rough drafts, book chapters) into Douglas Adams' distinctive style: deadpan British absurdism, cosmic-scale bathos, philosophical tangents that arrive at profound irrelevance, and sentences that hang in the air much the same way that bricks don't.
When to Use
- News articles needing absurdist transformation
- Technical/academic writing requiring comic deflation
- Book chapters or stories needing Adams voice
- Any text where user requests "Douglas Adams style" or "Hitchhiker's style"
- Parody requests for sci-fi or technical subjects
- User wants to add British absurdist humor
When NOT to Use
- Already comedic text in different style (Pratchett, Wodehouse)
- Serious content requiring gravity (eulogies, medical instructions)
- User explicitly wants different humor style
- Very short snippets (<100 words) unless expanding to full piece
Core Style Principles
The Adams Formula
Setup → Elevation → Bathos → Recovery → Deeper Absurdity
Every paragraph follows this arc: establish something seemingly important, elevate reader expectations, deflate with anti-climax, briefly recover, then reveal the deflation was itself setup for greater absurdity.
Signature Techniques
- Cosmic Bathos - Universe-scale importance deflated to trivial concern
- The Brick Simile - Comparisons to what things DON'T resemble
- Digressive Tangents - Parenthetical asides that become the point
- Deadpan Delivery - Ludicrous statements in matter-of-fact tone
- Philosophical Anti-Climax - Deep questions answered with mundane observations
Execution Workflow
PRE-FLIGHT: Source Assessment (v0.9.0 - CRITICAL)
Before any transformation, assess the source:
- Get source text directly - Do NOT rely on WebFetch summaries for word count
- Count source words accurately
- If >10,000 words: Use chunking workflow
python scripts/chunk_text.py --input source.txt --output-dir chunks/
CRITICAL WARNING: WebFetch summarizes long web pages. If transforming a URL:
- Fetch the full HTML/text first
- Count words in the FULL source
- If >10,000 words, chunk before processing
- Never transform a summary thinking it's the full text
Step 1: Content Extraction & Fresh Material (CRITICAL)
Run Phase 0 BEFORE any transformation.
Step 1a: Content Extractor
- Prompt:
prompts/content_extractor.md
- Measures source length and sets target range (60% to 115% of source)
- Extracts ALL key points with importance ratings (HIGH/MEDIUM/LOW)
- Maps structure for section-by-section transformation
LENGTH TARGETS
==============
Source: [N] words
Minimum (floor): [N × 0.60] words - DO NOT GO BELOW
Maximum (ceiling): [N × 1.15] words
Ideal: [N × 1.00] to [N × 1.10] words
Step 1b: Originality Engines (run in parallel)
Observation Engine (prompts/observation_engine.md):
- Questions assumptions in source text
- Generates FRESH observations specific to THIS topic
- Key method: "What's actually strange about this if you think about it?"
- Output: Fresh Observation Inventory (Grade A/B/C)
Contemporary Adams Vision (prompts/contemporary_adams_vision.md):
- Identifies 2015-2025 absurdities relevant to topic
- Modern equivalents for digital watches, Vogon bureaucracy, etc.
- Optional web search for recent examples (last 2 years prioritised)
- Output: Contemporary angles for tangents/bathos
Combined Output: Fresh Material Inventory - travels through entire pipeline.
The Preserve-First Principle: Adams added tangents ON TOP of content. He didn't compress. The Adamiser must do the same.
The Originality Principle: Adams invented new jokes. "Hordes of writers have tried to copy his style... but none successfully." We use his TECHNIQUES, not his specific jokes.
Step 2: Plan Transformation
Produce 5-bullet strategy:
- Opening approach (cosmic scale or mundane zoom-in)
- Primary bathos targets (what to deflate)
- Digression opportunities (tangent subjects)
- Simile/metaphor inversions planned
- Closing anti-climax structure
Step 3: Parallel Agent Analysis
CRITICAL: Launch these 4 specialist agents IN PARALLEL using a single message with multiple Task tool calls:
Agent 1: Bathos Architect
- Prompt: prompts/bathos_architect.md
- Task: Identify serious statements, map bathos insertion points
- Output: Bathos insertion map with patterns (A-D) and locations
Agent 2: Metaphor Engineer
- Prompt: prompts/metaphor_engineer.md
- Task: Inventory existing comparisons, suggest new ones using domain taxonomy
- Output: Comparison inventory and insertion suggestions (Mechanical/Animal/Human/Abstract/Natural)
Agent 3: Rhythm Analyst
- Prompt: prompts/rhythm_analyst.md
- Task: Audit sentence/paragraph structure against corpus targets
- Output: Rhythm map with correction recommendations (30% short, 32% single-sentence paragraphs)
Agent 4: Dialogue Transformer (if dialogue present)
- Prompt: prompts/dialogue_transformer.md
- Task: Audit dialogue tags (90% "said" rule), structure exchanges
- Output: Tag corrections and structural recommendations
Wait for all agents to complete, then synthesize outputs following priority order: Rhythm → Bathos → Metaphor → Dialogue.
Step 4: Transform (Multi-Phase)
Phase A: Opening
Apply opening system from reference/adams_style_guide.md:
- Cosmic Zoom: Start at universal scale, crash to trivial
- Mundane Enormity: Begin with boring detail, reveal cosmic significance
- Authoritative Absurdity: Guide-entry format with deadpan impossibility
Phase B: Body Transformation
For each paragraph:
- Identify the "serious point"
- Elevate its apparent importance
- Insert digressive tangent (see
reference/digression_systems.md)
- Apply bathos formula (see
reference/bathos_patterns.md)
- Weave sentence architecture (see
reference/sentence_architecture.md)
Phase C: Closing
Apply triple-layer Adams closing:
- Layer 1: Return to original subject as if nothing absurd occurred
- Layer 2: Casual mention of universe-ending implication
- Layer 3: Character worrying about tea/towels/bureaucracy
Step 5: Four-Layer Quality Review (Ultrathink)
This is the critical differentiator: Style fidelity alone produces parody. Originality prevents imitation. Meaning amplification produces excellence.
Run sequentially - each layer feeds the next:
Layer 1: Style Fidelity Check
- Prompt:
prompts/layer1_fidelity_check.md
- Input: First draft from Step 4
- Task: DETECT (not fix) style deviations using corpus-calibrated metrics
- Output: Fidelity report with specific flags (target: ≥6/8 checks passed)
Anti-Cliché Layer (NEW)
- Prompt:
prompts/anti_cliche_layer.md
- Input: First draft + Layer 1 report
- Task: Detect recycled Adams tropes and insufficient joke depth
- Checks:
- BANNED (0 uses): brick similes, 42, Vogons, Babel fish, "Don't Panic"
- LIMITED (budget per 2000 words): "not entirely unlike", tea, towels
- "Twist more than once": Single-twist jokes flagged for deepening
- Output: Anti-Cliché report with replacement recommendations
Layer 2: Technique Refinement
- Prompt:
prompts/layer2_technique_refinement.md
- Input: First draft + Layer 1 report + Anti-Cliché report
- Task: Apply SURGICAL corrections + replace clichés with fresh material
- Output: Refined text + change log
Layer 3: Meaning Amplification
- Prompt:
prompts/layer3_meaning_amplification.md
- Input: Refined text + original source
- Task: Ensure Adams techniques AMPLIFY (not obscure) the message
- Output: Final text + amplification report
The Ultrathink Principle: Every Adams technique must serve the content's truth. Decorative mimicry fails Layer 3.
The "Twist More Than Once" Principle: Professional analysis notes "Douglas has to twist the line more than once." Single-layer jokes are flagged for deeper work.
See prompts/orchestrator.md for complete workflow coordination.
Step 6: Validate
MANDATORY: Run validate_metrics.py after Phase 2 synthesis:
python scripts/validate_metrics.py --source original.txt --output transformed.txt --json
For long texts (>10,000 words), use chunk_text.py FIRST:
python scripts/chunk_text.py --input long_source.txt --output-dir chunks/
Script checks (deterministic):
- Word count ratio (60-115% of source)
- Banned phrases (Adams quotes, clichés, book report patterns, PowerPoint headers)
- Deadpan markers ("in fact"/"of course" density)
- Sentence distribution
- Exclamation marks (<1 per 2000 words)
- Question frequency (8-10 per 1000 words)
Optional: Run validate_adams_style.py for comprehensive final check:
python scripts/validate_adams_style.py <output_file>
Additional checks:
- Bathos presence (varied placement, NOT formulaic intervals)
- Digression density (at least 1 tangent per 300 words)
- Inverted comparisons ("not like X, but like Y")
- Sentence rhythm (30% short ≤10 words, 31% single-sentence paragraphs)
- Adams vocabulary markers ("quite"/"rather" > "utterly")
- Dialogue tag distribution (90% "said")
Step 7: Report
Transformation Complete
**Source Analysis:**
- Original style: [formal/journalistic/technical]
- Original length: [X words]
- Bathos opportunities identified: [Y]
**Output Specifications:**
- Final length: [X words]
- Bathos instances: [N]
- Digressions inserted: [M]
- Simile inversions: [K]
- Validation: [PASS/FAIL with details]
**Key Transformations:**
1. [e.g., Converted Q3 earnings report to Vogon poetry analysis]
2. [e.g., Inserted 400-word tangent on digital watches]
3. [e.g., Applied cosmic bathos to meeting agenda]
CRITICAL: Contextual Mode Detection
Do not apply global averages. Adams wrote differently in different contexts. Before any transformation, segment text by mode:
| Mode | Short Sentences | Tangent Density | Reluctance |
|---|
| Guide Entry | 20-25% | LOW | LOW (confident absurdity) |
| Narrative Action | 35-40% | MEDIUM | HIGH for Arthur, LOW for Ford |
| Philosophical Musing | 15-20% | HIGH (is the tangent) | MEDIUM (weary omniscience) |
| Dialogue Scene | 45-55% | LOW | Character-dependent |
| Descriptive Passage | 25-30% | MEDIUM-HIGH | MEDIUM (bored by magnificence) |
See reference/contextual_modes.md for full details.
The Marketing Copy Trap: AI tends toward efficiency. Adams was gloriously inefficient. If your output moves smoothly A→B→C, you've failed. Adams moved A→tangent→tangent→B.
Corpus-Validated Insights
Based on analysis of 528,808 words of authentic Adams text across 10 books:
Vocabulary (Often Imitated Wrong)
Top Intensifiers:
- "quite" (8.13/10k) and "rather" (6.73/10k) are #1 and #2
- "utterly" is OVERUSED in imitations (only 0.96/10k in real Adams)
Sentence Distribution:
- 30% short (≤10 words) - MORE punchy than commonly assumed
- 46% medium (11-30 words)
- 15% long (31-50 words)
- 8% very long (>50 words)
Questions: ~9 per 1000 words (high - drives philosophical voice)
Signature "brick simile": Use sparingly - only 0.15 per 10k words
Opening Patterns (NEW)
| Type | Frequency | Example |
|---|
| Cosmic Zoom | 33% | "Far out in the uncharted backwaters..." |
| Character Action | 25% | "The regular early morning yell of horror..." |
| Philosophical | 17% | "There is a theory which states..." |
| Meta-Commentary | 17% | "The history of the Galaxy has got muddled..." |
Bathos Timing is NOT Formulaic (NEW)
Real Adams places first bathos anywhere from 30 words to 2,200 words into a piece. The variation IS the pattern. Do not apply bathos at fixed intervals.
Dialogue (NEW)
- Use "said" for 90% of dialogue tags
- Reserve "burbled" for machines, "rasped/rumbled" for authority figures
- ~31.9% of paragraphs are single sentences (use for punchlines)
Metaphor Patterns (NEW)
- 693 comparisons across corpus: 37.5% "like a/the", 28.3% "as if"
- Inverted comparisons ("not like X, but like Y") are signature technique
- Categories: Mechanical (34%), Animal (28%), Human (22%), Abstract (10%)
Chapter Endings (NEW)
- End with deflation/bathos (35%), dialogue punchline (30%), physical comedy (20%)
- Final sentences typically SHORT (3-15 words)
- Recurring: "Or so it would seem", "It was that simple"
See reference/literary_analysis.md for complete analysis and reference/corpus_analysis_summary.md for statistics.
Reference Files
Agent Prompts (Multi-Layer System)
Phase 0 - Extraction & Fresh Material:
prompts/content_extractor.md - RUN FIRST - Extracts key points, sets length targets
prompts/observation_engine.md - NEW - Generates FRESH observations by questioning assumptions
prompts/contemporary_adams_vision.md - NEW - 2015-2025 absurdities, modern digital watch equivalents
Phase 1 - Parallel Analysis:
prompts/orchestrator.md - Coordinates entire workflow with parallel agents and 4-layer review
prompts/bathos_architect.md - Bathos analysis and insertion (4 patterns, corpus timing)
prompts/metaphor_engineer.md - Comparison crafting (5 techniques, domain taxonomy)
prompts/rhythm_analyst.md - Sentence/paragraph structure (corpus targets)
prompts/dialogue_transformer.md - Dialogue tags and exchange patterns
prompts/tangent_injector.md - Tangent insertion using fresh material
Phase 3 - Quality Layers:
prompts/layer1_fidelity_check.md - Style deviation detection (12 checkpoints)
prompts/anti_cliche_layer.md - NEW - Trope detection, originality verification, "twist more than once"
prompts/layer2_technique_refinement.md - Surgical corrections + cliché replacement
prompts/layer3_meaning_amplification.md - ULTRATHINK - Ensures techniques serve message
Core References
reference/contextual_modes.md - CRITICAL - The 5 contextual modes with mode-specific targets (Guide Entry, Narrative, Philosophical, Dialogue, Descriptive)
reference/original_patterns.md - NEW - 100+ pre-generated Adams-esque patterns (bathos, similes, tangents, etc.) - use instead of recycling Adams' specific jokes
reference/literary_analysis.md - Comprehensive literary patterns from 10-book corpus study
reference/corpus_analysis_summary.md - Real statistics from Adams corpus
reference/adams_style_guide.md - Complete style breakdown with examples
reference/bathos_patterns.md - 12 bathos formulas with templates
reference/sentence_architecture.md - Adams-specific sentence patterns
reference/digression_systems.md - Tangent insertion techniques
reference/vocabulary_guide.md - Adams vocabulary, neologisms, phrase patterns
Scripts
Primary (Integrated into Pipeline):
scripts/validate_metrics.py - MANDATORY - Deterministic validation (banned phrases, word counts, deadpan markers). Run after Phase 2, output feeds Phase 3 layers. Use --json for programmatic output.
scripts/chunk_text.py - FOR LONG TEXTS - Splits texts >10,000 words into ~3000-word chunks at natural breaks for parallel processing.
Optional Standalone Tools:
scripts/validate_adams_style.py - Comprehensive style validation (can run independently for final check)
scripts/bathos_detector.py - Detect and score anti-climax patterns (exploratory analysis)
scripts/vocabulary_checker.py - Check Adams vocabulary markers (exploratory analysis)
scripts/analyse_corpus.py - Analyse text for Adams style statistics (reference tool)
Examples
examples/news_to_adams.md - News article transformation
examples/academic_to_adams.md - Academic paper transformation
examples/technical_to_adams.md - Technical doc transformation
Templates
templates/guide_entry.md - Hitchhiker's Guide entry format
templates/chapter_opening.md - Novel chapter opening format
templates/cosmic_zoom.md - Universal-to-trivial structure
Output Requirements
Must include:
- Bathos instances (vary placement - NOT at fixed intervals, range 30-2200 words to first)
- At least 1 digressive tangent per 300 words
- At least 1 inverted comparison ("not like X, but like Y") per 500 words
- Mode-appropriate short sentence percentage (see contextual modes)
- ~31% single-sentence paragraphs for rhythm and punchlines
- Use "quite"/"rather" NOT "utterly" for intensification
- Use "said" for 90% of dialogue tags
Originality Requirements (v0.9.0 EXPANDED):
- 0 BANNED Adams tropes (brick similes, 42, Vogons, Babel fish, "Don't Panic", whale/petunias)
- 0 VERBATIM Adams openings ("Far out in the uncharted backwaters...", "In the beginning the Universe was created...")
- 0 Book report constructions ("the letter mentions", "the author argues", "in the article")
- 0 PowerPoint headers (ALL CAPS section breaks - use narrative bridges instead)
- LIMITED tropes within budget (max 1 "not entirely unlike" per 2000 words)
- ≥70% of jokes must have multiple twists ("twist more than once")
- Fresh observations from Observation Engine, not recycled Adams material
- Contemporary (2015-2025) references where relevant
Length: 60% to 115% of source (target: 100-110%)
- Below 60%: Content has been lost - FAIL
- 100-110%: Ideal (preserved content + Adams additions)
- Above 115%: Review for over-tangenting
Content Coverage: ≥90% of source key points must appear in output
- HIGH importance points: 100% required (non-negotiable)
- MEDIUM importance points: 85%+ required
- Layer 3 verifies coverage before approval
Format: Markdown (or HTML using appropriate template)
Endings: Short final sentence (3-15 words), deflation or dialogue punchline
Invocation: Provide source text and say "Adamise this", "Douglas Adams style", or "make this absurd" - skill auto-detects from context.
Modes: "plan only" for strategy preview, "validate only" to check existing text, "guide entry" for Hitchhiker's format specifically.
Audio Output (Optional)
Generate audio narration of Adamised text using ElevenLabs v3:
ELEVENLABS_API_KEY="sk_da5157040e6b27a7d1eb9f73ab2b2891aaead490991d2bcf"
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/oaGwHLz3csUaSnc2NBD4" \
-H "xi-api-key: $ELEVENLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "YOUR_ADAMISED_TEXT", "model_id": "eleven_v3"}' \
--output adamised_audio.mp3
Voice: Boris (oaGwHLz3csUaSnc2NBD4)
Model: eleven_v3 (most expressive - sighs, laughs, whispers)
For long texts, split into chunks <5000 characters and concatenate audio files.