Agent Skills: Curses

Discover the structural costs hidden in your strengths through behavioral dimension analysis, strength-shadow extraction, and attitude recommendations.

UncategorizedID: jongwony/epistemic-protocols/curses

Install this agent skill to your local

pnpm dlx add-skill https://github.com/jongwony/epistemic-protocols/tree/HEAD/epistemic-cooperative/skills/curses

Skill Files

Browse the full folder contents for curses.

Download Skill

Loading file tree…

epistemic-cooperative/skills/curses/SKILL.md

Skill Metadata

Name
curses
Description
"Discover the structural costs hidden in your strengths through behavioral dimension analysis, strength-shadow extraction, and attitude recommendations."

Curses

Discover the structural costs hidden in your strengths.

Every strength casts a shadow. The shadow is not a flaw — it is the structural cost of a capability. Understanding the cost transforms a curse into a conscious trade-off.

When to Use

Invoke this skill when:

  • Discovering structural costs hidden in your strengths
  • Analyzing behavioral patterns for self-improvement recommendations
  • Generating attitude principles and practice matrix
  • Reflective questions about working patterns and their trade-offs

Skip when:

  • Exploring philosophical tradition match (use /sophia instead)
  • Quick single-protocol question (answer directly)
  • No session history exists and user prefers manual exploration

Pipeline

| Phase | What | Mode | |-------|------|------| | 1. Collect | Gather behavioral data | dimension-profiler agent | | 2. Analyze | Strength-Shadow extraction | AI + user dialogue | | 3. Recommend | Attitude principles + practice matrix | AI proposes | | 4. Report | Generate HTML report | Automated |

If the user provides a specific question (e.g., "What are my curses?"), orient the analysis toward that question.


Phase 1: Data Collection

Same-session reuse: If dimension-profiler output is already available in this conversation (from a prior /sophia or /curses run), skip Phase 1 entirely and reuse that output. Both skills produce identical profiler results.

Two-step delegation (same pipeline as /sophia):

Step 1: Run coverage-scanner agent (see agents/coverage-scanner.md) to get pre-aggregated session data (protocol counts, friction, session types, tools).

Step 2: Pass coverage output to dimension-profiler agent (see agents/dimension-profiler.md):

Analyze this user's behavioral dimensions from their session data.

coverage_data: [paste coverage-scanner output here]

data_sources:
  rules_dir: ~/.claude/rules/
  claude_md: ~/.claude/CLAUDE.md
  settings_json: ~/.claude/settings.json

data_context: session-enriched

Return the dimension profile table with scores, confidence, and raw signals.

When coverage_data is provided, omit sample_size — the profiler derives dimensions from aggregate data and does not sample raw files.

If a dimension's confidence is "low", include it in the analysis but mark it as provisional and note this in the report.


Phase 2: Strength-Shadow Analysis

From the dimension profile, identify strengths and their structural costs.

Extraction method

For each dimension scoring above 65 (or below 35 — extremes in either direction):

  1. Name the strength: What capability does this extreme enable?
  2. Find the shadow: What structural cost does this extreme create?
  3. Identify the mechanism: Why does this strength produce this specific cost?
  4. Rate severity: structural (inherent), recurring (frequent), or conditional (context-dependent)
  5. Cite evidence: Link to specific data from the dimension profiler

Common strength-shadow patterns

These are heuristic starting points, not fixed outputs. Adapt based on actual data.

| Dimension extreme | Strength | Shadow | |-------------------|----------|--------| | D2 high (Doubt) | Catches errors early | Verification depth becomes opportunity cost | | D4 high (Systematic) | Consistent governance | Rule accumulation creates complexity | | D5 high (UU) | Discovers new patterns | May defer KK maintenance | | D6 high (Extended Mind) | Effective delegation | Curse activates on delegation failure | | D1 high (Abductive) | Creative hypothesis | May skip systematic validation | | D3 high (Dialogical) | Deep understanding | Extended exchanges consume time |

Cross-dimensional patterns

Look for patterns that emerge from dimension COMBINATIONS:

  • D2 high + D4 high: "The cure-as-disease pattern" — each verification failure produces a new rule, which accumulates
  • D5 high + D6 high: "Extended Mind strategy" — UU preference + AI delegation may be a strategy, not a curse (validate with user)
  • D1 high + D2 high: "Bold conjecture + rigorous refutation" — Popperian pattern, strong if balanced

Dual-interpretation guidance (cold-start awareness)

When presenting strength-shadow pairs, some combinations may be either a curse OR a deliberate strategy. In context-rich sessions, the user may have already articulated this distinction. In cold-start sessions, the AI must proactively surface both interpretations before the user validates.

Patterns that require dual-interpretation:

  • D5 high + D6 high: Could be "deliberate Extended Mind strategy with quality bridges" OR "KK neglect via over-delegation"
  • D2 high + D4 high: Could be "systematic verification infrastructure that scales" OR "cure-as-disease accumulation"

For these dual-interpretation patterns specifically, retain a Constitution gate before downstream derivation: "This pattern admits two readings — [strategy interpretation] or [curse interpretation]. Which better describes your experience?" The user's intent here is project-profile category (a) — user IS the measurement target — and cannot be auto-resolved regardless of profile; downstream recommendations depend on this choice. Single-interpretation pairs (e.g., D1+D2, isolated dimensions) bypass this gate and proceed via the relay path described under "User dialogue" below.

User dialogue

When presenting dimensions to the user, always include the human-readable explanation from the dimension-profiler output (e.g., "D4 Rule Orientation — how you govern work") so users unfamiliar with the framework understand what each dimension measures.

Present single-interpretation strength-shadow pairs (the majority — those without the dual-interpretation gate above) as text output and proceed directly to Phase 3 recommendations. End the Phase 2 output with a visible red-line discovery line so the correction pathway is explicit: "If any pair seems misclassified, say so — I'll re-derive from there." The user may red-line via free response at any subsequent turn — confirm, reframe, dismiss, or add context. Counter-evidence that changes the structural category triggers re-derivation of downstream recommendations on the next turn.


Phase 3: Recommendations

From validated strength-shadow pairs, derive:

Attitude principles (max 4)

Each principle addresses a specific shadow:

Principle N — [Title]

[2-3 sentence explanation of the principle and why it addresses this shadow]

Application:
[Concrete, actionable guidance for daily practice]

Rank by ROI — which principle would have the highest impact if adopted? Mark the highest-ROI principle explicitly.

Practice matrix

Map principles to concrete situations:

| Situation | Principle | Action | Trigger | |-----------|-----------|--------|---------| | When X happens | Principle N | Do Y | Z condition |

Include 4-6 rows covering the most common situations.


Phase 4: Report Generation

Generate an HTML report following the cooperative's design system.

Design system source

Read one of:

  • ~/.claude/usage-data/report.html — extract CSS
  • Cooperative's skills/report/references/html-template.md — use as template basis
  • skills/curses/references/report-template.md — curses-specific components

Context awareness

Check the dimension-profiler's Data Context field:

  • session-enriched: Report may reference protocol chaining results, prior /sophia output, or session-specific observations. Include these in relevant sections.
  • data-only: Report is generated purely from behavioral data. Keep analysis grounded in the dimension scores and raw signals.

Mark the report subtitle with the context tier (e.g., "708 sessions | data-only" or "708 sessions | session-enriched").

Required sections

  1. At a Glance — 3-4 bullet summary with section links
  2. Dimension Profile — 6 horizontal bars with scores and human-readable explanation per dimension
  3. Strengths — Green cards with evidence and dimension tag
  4. Structural Costs — White cards with severity badge and mitigation
  5. Attitude Recommendations — Gradient cards ranked by ROI
  6. Practice Matrix — Situation to principle to action table
  7. Health Indicators — Green/yellow/red dots for monitored vs unmonitored areas
  8. Next Steps — 2-3 concrete horizon cards

Optional sections (if /sophia was run in same session)

If the dimension profile and philosopher match are available from a prior /sophia run in this session, include:

  • Philosophical Identity — 2x2 grid of philosophy cards
  • Division of Labor — Human-AI role visualization

Output

Save to ~/.claude/usage-data/curses-profile.html Open in browser: open <filepath>


Edge Cases

  • New user (<5 sessions): Analyze rules only. Present as "Configuration-based profile — behavioral data will improve accuracy over time."
  • No extreme dimensions (all 35-65): "Your profile is notably balanced. This is itself a strength (Aristotelian phronesis) with its own shadow: you may lack the specialization that comes from extreme focus."
  • User reframes curse as strategy: Accept the reframe. Update the strength-shadow pair to reflect the new category. Re-derive recommendations from the updated structure.
  • Specific question provided: Orient the entire analysis toward answering that question. The report sections should reflect the specific inquiry.
Curses Skill | Agent Skills