intent-sink

Intent Sink Skill

16 stars

Best use case

intent-sink is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

Intent Sink Skill

Teams using intent-sink should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

  • You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

  • You only need a quick one-off answer and do not need a reusable workflow.
  • You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/intent-sink/SKILL.md --create-dirs "https://raw.githubusercontent.com/plurigrid/asi/main/plugins/asi/skills/intent-sink/SKILL.md"

Manual Installation

  1. Download SKILL.md from GitHub
  2. Place it in .claude/skills/intent-sink/SKILL.md inside your project
  3. Restart your AI agent — it will auto-discover the skill

How intent-sink Compares

Feature / Agentintent-sinkStandard Approach
Platform SupportNot specifiedLimited / Varies
Context Awareness High Baseline
Installation ComplexityUnknownN/A

Frequently Asked Questions

What does this skill do?

Intent Sink Skill

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

SKILL.md Source

# intent-sink Skill


> *"Where intents go to be validated. The final checkpoint before execution."*

## Overview

**Intent Sink** is the validation endpoint for intent-centric architectures. It validates that intents are well-formed, satisfiable, and safe before allowing execution.

## GF(3) Role

| Aspect | Value |
|--------|-------|
| Trit | -1 (MINUS) |
| Role | VALIDATOR |
| Function | Validates intents before execution |

## Architecture

```
┌─────────────────────────────────────────────────────────────────┐
│                      INTENT FLOW                                │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  User Intent    Solver       Intent Sink      Execution         │
│  (+1 GEN)      (0 COORD)     (-1 VAL)        (output)          │
│      │             │              │               │             │
│      ▼             ▼              ▼               ▼             │
│  ┌───────┐    ┌────────┐    ┌──────────┐    ┌─────────┐        │
│  │Declare│───►│ Solve  │───►│ Validate │───►│ Execute │        │
│  └───────┘    └────────┘    └──────────┘    └─────────┘        │
│                                  │                              │
│                                  ▼                              │
│                           ┌──────────┐                          │
│                           │ Reject ? │                          │
│                           └──────────┘                          │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘
```

## Validation Checks

```python
class IntentSink:
    """Final validation before intent execution."""

    TRIT = -1  # VALIDATOR role

    def validate(self, intent, solution):
        """Run all validation checks."""
        checks = [
            self.check_well_formed(intent),
            self.check_resource_conservation(solution),
            self.check_authorization(intent),
            self.check_deadlines(intent),
            self.check_slippage(intent, solution),
        ]
        return all(checks)

    def check_well_formed(self, intent):
        """Intent has valid structure."""
        required = ['type', 'constraints', 'deadline']
        return all(k in intent for k in required)

    def check_resource_conservation(self, solution):
        """Inputs balance outputs (no creation/destruction)."""
        input_sum = sum(r.quantity for r in solution.inputs)
        output_sum = sum(r.quantity for r in solution.outputs)
        return input_sum == output_sum

    def check_authorization(self, intent):
        """User authorized to create this intent."""
        return verify_signature(intent.signature, intent.user)

    def check_deadlines(self, intent):
        """Intent hasn't expired."""
        return intent.deadline > current_time()

    def check_slippage(self, intent, solution):
        """Solution meets slippage constraints."""
        if intent.type == 'swap':
            actual_rate = solution.output_amount / solution.input_amount
            min_rate = intent.min_rate * (1 - intent.slippage)
            return actual_rate >= min_rate
        return True
```

## Sink Modes

```python
class SinkMode(Enum):
    STRICT = "reject on any failure"
    LENIENT = "allow with warnings"
    DRY_RUN = "validate but don't execute"

class ConfigurableSink:
    def __init__(self, mode: SinkMode):
        self.mode = mode

    def process(self, intent, solution):
        result = self.validate(intent, solution)

        if self.mode == SinkMode.DRY_RUN:
            return {"valid": result, "executed": False}

        if not result and self.mode == SinkMode.STRICT:
            raise ValidationError("Intent failed validation")

        if not result and self.mode == SinkMode.LENIENT:
            log.warning(f"Intent {intent.id} has warnings")

        return {"valid": result, "executed": True}
```

## GF(3) Integration

```python
def intent_triad(intent, solver, sink):
    """
    Complete intent lifecycle with GF(3) conservation.

    intent (+1) + solver (0) + sink (-1) = 0 ✓
    """
    # Generation phase
    raw_intent = intent.declare()  # +1

    # Coordination phase
    solution = solver.solve(raw_intent)  # 0

    # Validation phase
    if sink.validate(raw_intent, solution):  # -1
        return solution.execute()
    else:
        return None

    # Net GF(3): +1 + 0 + (-1) = 0 ✓
```

## Anoma Integration

```juvix
-- Intent sink in Juvix
module IntentSink;

type ValidationResult :=
  | Valid : Solution -> ValidationResult
  | Invalid : Error -> ValidationResult;

validate : Intent -> Solution -> ValidationResult;
validate intent solution :=
  if (all-checks-pass intent solution)
    then Valid solution
    else Invalid (first-failure intent solution);

-- Compose with solver
process : Intent -> Maybe Transaction;
process intent :=
  case solve intent of
    | Nothing -> Nothing
    | Just solution ->
        case validate intent solution of
          | Valid s -> Just (execute s)
          | Invalid _ -> Nothing;
```

## GF(3) Triads

```
intent-sink (-1) ⊗ solver-fee (0) ⊗ anoma-intents (+1) = 0 ✓
intent-sink (-1) ⊗ dynamic-sufficiency (0) ⊗ polyglot-spi (+1) = 0 ✓
```

---

**Skill Name**: intent-sink
**Type**: Intent Validation
**Trit**: -1 (MINUS - VALIDATOR)
**GF(3)**: Final checkpoint for intent execution


## Scientific Skill Interleaving

This skill connects to the K-Dense-AI/claude-scientific-skills ecosystem:

### Graph Theory
- **networkx** [○] via bicomodule
  - Universal graph hub

### Bibliography References

- `general`: 734 citations in bib.duckdb

## Cat# Integration

This skill maps to Cat# = Comod(P) as a bicomodule in the Prof home:

```
Trit: 0 (ERGODIC)
Home: Prof (profunctors/bimodules)
Poly Op: ⊗ (parallel composition)
Kan Role: Adj (adjunction bridge)
```

### GF(3) Naturality

The skill participates in triads where:
```
(-1) + (0) + (+1) ≡ 0 (mod 3)
```

This ensures compositional coherence in the Cat# equipment structure.

Related Skills

juvix-intents

16
from plurigrid/asi

Juvix intent-centric language for Anoma with Geb compilation and GF(3) typed resources

anoma-intents

16
from plurigrid/asi

Anoma intent-centric architecture for cross-chain obstruction passing with Geb semantics and Juvix compilation

testing-android-intents-for-vulnerabilities

16
from plurigrid/asi

Tests Android inter-process communication (IPC) through intents for vulnerabilities including intent injection, unauthorized component access, broadcast sniffing, pending intent hijacking, and content provider data leakage. Use when assessing Android app attack surface through exported components, testing intent-based data flows, or evaluating IPC security. Activates for requests involving Android intent security, IPC testing, exported component analysis, or Drozer assessment.

jepsen-testing

16
from plurigrid/asi

Jepsen-style correctness testing for distributed systems under faults (partitions, crashes, clock skew) using concurrent operation histories and formal checkers (linearizability/serializability and Elle-style anomalies). Use when designing, implementing, or running Jepsen tests, or interpreting histories/violations.

Deterministic Color Generation via Metadata Hashing

16
from plurigrid/asi

**Status**: ✅ Production Ready

cyton-dongle

16
from plurigrid/asi

Connect and stream from OpenBCI Cyton/Daisy via USB dongle, including first-time radio channel pairing

asi-transient-agenda

16
from plurigrid/asi

Org-agenda-like transient views for ASI skill orchestration via nbb/squint + Emacs hydra

Topological Superintelligence (TSI)

16
from plurigrid/asi

Compositional AI framework using GF(3) triadic balance and category-theoretic foundations.

zx-calculus

16
from plurigrid/asi

Coecke's ZX-calculus for quantum circuit reasoning via string diagrams with Z-spiders (green) and X-spiders (red)

zulip-cogen

16
from plurigrid/asi

Zulip Cogen Skill 🐸⚡

zls-integration

16
from plurigrid/asi

zls-integration skill

zig

16
from plurigrid/asi

zig skill