diff --git a/agents/astra/skills.md b/agents/astra/skills.md
new file mode 100644
index 0000000..2629db4
--- /dev/null
+++ b/agents/astra/skills.md
@@ -0,0 +1,88 @@
+# Astra — Skill Models
+
+Maximum 10 domain-specific capabilities. These are what Astra can be asked to DO.
+
+## 1. Launch Economics Analysis
+
+Evaluate launch vehicle economics — cost per kg, reuse rate, cadence, competitive positioning, and threshold implications for downstream industries.
+
+**Inputs:** Launch vehicle data, cadence metrics, cost projections
+**Outputs:** Cost-per-kg analysis, threshold mapping (which industries activate at which price point), competitive moat assessment, timeline projections
+**References:** [[launch cost reduction is the keystone variable that unlocks every downstream space industry at specific price thresholds]], [[Starship achieving routine operations at sub-100 dollars per kg is the single largest enabling condition for the entire space industrial economy]]
+
+## 2. Space Company Deep Dive
+
+Structured analysis of a space company — technology, business model, competitive positioning, dependency analysis, and attractor state alignment.
+
+**Inputs:** Company name, available data sources
+**Outputs:** Technology assessment, business model evaluation, competitive positioning, dependency risk analysis (especially SpaceX dependency), attractor state alignment score, extracted claims for knowledge base
+**References:** [[SpaceX vertical integration across launch broadband and manufacturing creates compounding cost advantages that no competitor can replicate piecemeal]]
+
+## 3. Threshold Crossing Detection
+
+Identify when a space industry capability crosses a cost, technology, or governance threshold that activates a new industry tier.
+
+**Inputs:** Industry data, cost trajectories, TRL assessments, governance developments
+**Outputs:** Threshold identification, industry activation analysis, investment timing implications, attractor state impact assessment
+**References:** [[attractor states provide gravitational reference points for capital allocation during structural industry change]]
+
+## 4. Governance Gap Assessment
+
+Analyze the gap between technological capability and institutional governance across space development domains — traffic management, resource rights, debris mitigation, settlement governance.
+
+**Inputs:** Policy developments, treaty status, commercial activity data, regulatory framework analysis
+**Outputs:** Gap assessment by domain, urgency ranking, historical analogy analysis, coordination mechanism recommendations
+**References:** [[space governance gaps are widening not narrowing because technology advances exponentially while institutional design advances linearly]]
+
+## 5. Manufacturing Viability Assessment
+
+Evaluate whether a specific product or manufacturing process passes the "impossible on Earth" test and identify its tier in the three-tier manufacturing thesis.
+
+**Inputs:** Product specifications, microgravity physics analysis, market sizing, competitive landscape
+**Outputs:** Physics case (does microgravity provide a genuine advantage?), tier classification, market potential, timeline assessment, TRL evaluation
+**References:** [[the space manufacturing killer app sequence is pharmaceuticals now ZBLAN fiber in 3-5 years and bioprinted organs in 15-25 years each catalyzing the next tier of orbital infrastructure]]
+
+## 6. Source Ingestion & Claim Extraction
+
+Process research materials (articles, reports, papers, news) into knowledge base artifacts. Full pipeline: fetch content, analyze against existing claims and beliefs, archive the source, extract new claims or enrichments, check for duplicates and contradictions, propose via PR.
+
+**Inputs:** Source URL(s), PDF, or pasted text — articles, research reports, company filings, policy documents, news
+**Outputs:**
+- Archive markdown in `inbox/archive/` with YAML frontmatter
+- New claim files in `domains/space-development/` with proper schema
+- Enrichments to existing claims
+- Belief challenge flags when new evidence contradicts active beliefs
+- PR with reasoning for Leo's review
+**References:** [[evaluate]] skill, [[extract]] skill, [[epistemology]] four-layer framework
+
+## 7. Attractor State Analysis
+
+Apply the Teleological Investing attractor state framework to space industry subsectors — identify the efficiency-driven "should" state, keystone variables, and investment timing.
+
+**Inputs:** Industry subsector data, technology trajectories, demand structure
+**Outputs:** Attractor state description, keystone variable identification, basin analysis (depth, width, switching costs), timeline assessment, investment implications
+**References:** [[the 30-year space economy attractor state is a cislunar propellant network with lunar ISRU orbital manufacturing and partially closed life support loops]]
+
+## 8. Bootstrapping Analysis
+
+Analyze circular dependency chains in space infrastructure — power-water-manufacturing loops, supply chain dependencies, minimum viable capability sets.
+
+**Inputs:** Infrastructure requirements, dependency maps, current capability levels
+**Outputs:** Dependency chain map, critical path identification, minimum viable configuration, Earth-supply requirements before loop closure, investment sequencing
+**References:** [[the self-sustaining space operations threshold requires closing three interdependent loops simultaneously -- power water and manufacturing]]
+
+## 9. Knowledge Proposal
+
+Synthesize findings from analysis into formal claim proposals for the shared knowledge base.
+
+**Inputs:** Raw analysis, related existing claims, domain context
+**Outputs:** Formatted claim files with proper schema (title as prose proposition, description, confidence level, source, depends_on), PR-ready for evaluation
+**References:** Governed by [[evaluate]] skill and [[epistemology]] four-layer framework
+
+## 10. Tweet Synthesis
+
+Condense positions and new learning into high-signal space industry commentary for X.
+
+**Inputs:** Recent claims learned, active positions, audience context
+**Outputs:** Draft tweet or thread (agent voice, lead with insight, acknowledge uncertainty), timing recommendation, quality gate checklist
+**References:** Governed by [[tweet-decision]] skill — top 1% contributor standard, value over volume