Back to Agents That Account for Themselves
system2026Running system and technical note

A Typed Algebra for Agent Skills

The typed skill algebra turns a directory of prompts into a composable system. Skills are categorized by what they do, checked by a linter, and routed by type compatibility so chains can be reasoned about before execution.

Paper workspace

Live draft structure

live-draft

Artifacts

System source mapped

This is currently a running-system page, not a rendered paper.

source-only

Editable source

Running system summary. No standalone manuscript found in this pass; this should become a technical note with evidence from the skills corpus and linter.

Source anchors

Codex/Claude skills corpus

skill linter and typed skill registry

blog/type-system-for-skills

Method tags

typed skillscomposition algebralinterrouting

Ingest intersections

skillstypesagentscompositionlinting

Status

Running; 99.7% of the library typed, linter enforced.

Key claims

01

A skill library without types is hard to compose safely.

02

Generators, transformers, reducers, distributors, effectors, and auditors form a practical operational algebra.

03

Linting skill composition prevents some classes of agent drift.

Public reading note

System summary public; exact internal skill corpus is private.

Standard skeleton

What this paper must keep proving

Schema

problem

A large skill library becomes ungovernable if every skill is just text with trigger phrases.

method

Assign operational types to skills and check composition before routing or chaining.

implementation

Generator, transformer, reducer, distributor, effector, and auditor categories plus linter/router enforcement.

data

The internal skill corpus and routing outcomes; public release should summarize counts and examples without leaking private operations.

evaluation

Type coverage, linter rejection quality, routing success, and failure reduction in composed workflows.

references

Typed workflow systems, category-inspired composition, tool-use agents, skill libraries.

openQuestions

The next step is a concise technical note with measured linter and routing outcomes.

Checkpoints and references

Proof chain

paperpending

Claim checkpoint

central-claim slot

Every central claim must point to a proof anchor or remain labeled as speculative.

implementationpending

Implementation checkpoint

implementation-map slot

Every method should identify the code path, harness, schema, or protocol that embodies it.

experimentpending

Evidence checkpoint

evidence-manifest slot

Every reported result should point to run IDs, packet IDs, data snapshots, commits, or review artifacts.

external-referencepending

Reference checkpoint

references slot

Every external claim should resolve to a cited paper, benchmark, standard, or documented prior system.

paperpending

Release checkpoint

release-gate slot

Every PDF needs a named condition before it can move from draft to citation-ready.