constructs

name	constructs
description	Browse and install construct packs from the Loa Constructs Registry
role	implementation
allowed-tools	Read, Grep, Glob, WebFetch, Bash(gh repo *), Bash(gh release *)
capabilities	{"schema_version":1,"read_files":true,"search_code":true,"write_files":false,"execute_commands":{"allowed":[{"command":"gh","args":["repo","*"]},{"command":"gh","args":["release","*"]}],"deny_raw_shell":true},"web_access":true,"user_interaction":false,"agent_spawn":false,"task_management":false}
cost-profile	moderate

Browsing Constructs Skill

Purpose

Unified construct discovery surface for the Constructs Network. This skill is a thin API client — all search intelligence, ranking, and composability analysis lives in the Constructs Network API.

API Configuration

• Base URL: LOA_CONSTRUCTS_API_URL env var, default https://api.constructs.network/v1
• Auth: LOA_CONSTRUCTS_API_KEY env var or ~/.loa/credentials.json (for premium packs)

Invocation

/constructs (no args) — Browse-First Discovery

Fetch the full catalog and display grouped by category.

API call:

GET {API}/constructs?per_page=100&include_composability=true

Output format:

Constructs Network — 23 constructs

  design (8)
    🎨 artisan (14 skills) — Design systems craft
    🖼️ the-easel (4 skills) — Aesthetic direction studio
    ⚒️ the-mint (8 skills) — Material transformation pipeline

  security (3)
    🛡️ hardening (7 skills) — Defensive artifact forge
    🧪 crucible (5 skills) — Journey validation testing
    🔑 dynamic-auth (3 skills) — Wallet identity resolution

  research (2)
    🕳️ k-hole (2 skills) — Non-extractive pair research
    🦎 gecko (3 skills) — Autonomous ecosystem intelligence

  ...

Browse: constructs.network/constructs

Grouping rules:

1. Group constructs by category field from API response
2. Sort categories alphabetically
3. Within each category, sort by skill count descending
4. Show category name with construct count in parens
5. Each construct line: icon + slug + (N skills) + em-dash + short description

---

/constructs <natural query> — Intent Search

Fetch ranked results matching the user's natural language query.

API call:

GET {API}/constructs?q={query}&include_skills=true&include_composability=true

Output format:

Constructs Network — 23 constructs

Matching "security audit":

  🛡️ hardening (7 skills) [VERIFIED]
     Defensive artifact forge
     Skills: /threat-model  /security-review  /pen-test  /postmortem
     Works with: observer → protocol
     Install: /constructs install hardening

  🧪 crucible (5 skills) [COMMUNITY]
     Journey validation testing
     Skills: /validate  /test-journey  /detect-gaps
     Works with: observer ↔ (circular)
     Install: /constructs install crucible

Browse: constructs.network/constructs?q=security+audit

Display rules:

1. Show results in API-returned rank order (most relevant first)
2. Include verification tier in brackets: [VERIFIED], [COMMUNITY], [UNVERIFIED]
3. List up to 4 skill commands inline (truncate with ... if more)
4. Show composability edges using arrows: → (depends on), ← (depended by), ↔ (circular)
5. Include install command for each result

Zero results: If the API returns no matches, output:

No constructs match "<query>".
Try: npx skills find <query>

---

/constructs compose — Composability View

Show what constructs pair with currently installed constructs.

Steps:

1. Detect installed constructs:

   installed=$(.claude/scripts/constructs-loader.sh list 2>/dev/null)
   

2. Fetch all constructs with composability:

   GET {API}/constructs?per_page=100&include_composability=true
   

3. Filter to show only constructs that have composability edges with installed ones

Output format:

Composability for your installed constructs:

  You have: observer, artisan

  Works with observer:
    → crucible (required, circular) — Journey validation testing
    → hardening (depends on observer) — Defensive artifact forge
    → protocol (depends on observer) — On-chain verification

  Works with artisan:
    → the-easel (depends on artisan) — Aesthetic direction studio
    → the-speakers (optional) — Psychoacoustic engineering

  Install: /constructs install <slug>

Edge cases:

• No constructs installed: "No constructs installed. Run /constructs to browse."
• Installed constructs have no composability edges: "Your constructs have no declared composability relationships."

---

/constructs install <slug> — Install

Direct installation without browse UI.

1. Validate pack slug provided
2. Run: .claude/scripts/constructs-install.sh pack <slug>
3. Report result with installed skill commands

---

/constructs register — Register

Register a new construct with the network. Delegates to existing registration flow.

---

/constructs sync — Sync

Sync installed constructs with the registry. Delegates to existing sync flow.

---

/constructs status — Status

Show status of installed constructs, versions, and update availability.

.claude/scripts/constructs-loader.sh list
.claude/scripts/constructs-loader.sh check-updates

---

/constructs auth — Authentication

Check or set up authentication for premium packs.

auth (no args) — Check Status

.claude/scripts/constructs-auth.sh status

auth setup — Configure API Key

Guide user through API key setup. Keys from: LOA_CONSTRUCTS_API_KEY env var, ~/.loa/credentials.json, or ~/.loa-constructs/credentials.json (legacy).

.claude/scripts/constructs-auth.sh setup <api_key>

Passive Triggers

This skill should activate (with search mode) when the user says things like:

• "find a skill for ..."
• "find a construct ..."
• "I need help with ..."
• "is there a construct that ..."
• "what construct can ..."
• "search constructs ..."

When a passive trigger fires, treat the trailing text as the search query and execute the /constructs <query> flow.

API Response Shape

The skill expects the API to return constructs in this shape:

{
  "data": [
    {
      "slug": "hardening",
      "name": "Hardening",
      "description": "Defensive artifact forge",
      "icon": "🛡️",
      "category": "security",
      "skills_count": 7,
      "verification_tier": "VERIFIED",
      "skills": [
        { "slug": "threat-model", "command": "/threat-model" }
      ],
      "composability": {
        "depends_on": ["observer"],
        "depended_by": ["protocol"],
        "optional": []
      }
    }
  ],
  "meta": {
    "total": 23,
    "page": 1,
    "per_page": 100
  }
}

If the API returns a different shape, adapt gracefully. The skills and composability fields are only present when include_skills=true and include_composability=true query params are set.

Error Handling

Error	Handling
Network failure	Show error, suggest retrying. If cached data available, use it.
API returns 401	"Authentication required for this operation. Run /constructs auth setup."
API returns 404	"Construct not found. Run /constructs to browse available constructs."
API returns 500	"Constructs Network API error. Try again later."
No LOA_CONSTRUCTS_API_URL set	Use default https://api.constructs.network/v1

Per-Repo State

Installed packs go to .claude/constructs/packs/ which is gitignored.

Installation metadata tracked in .constructs-meta.json.

Related Scripts

• .claude/scripts/constructs-auth.sh — Authentication management
• .claude/scripts/constructs-browse.sh — Pack discovery (legacy, prefer API)
• .claude/scripts/constructs-install.sh — Installation
• .claude/scripts/constructs-loader.sh — Skill loading
• .claude/scripts/constructs-lib.sh — Shared utilities