Skip to content

Capabilities

Custom technical tools and CLIs

Focused command-line tools with an explicit contract, deterministic behaviour, and failure modes that can be tested.

Identification

Reference
L3D-CAP-002
Maturity
established
Applicable fields
Infrastructure & Operations, Reliability & Assurance
Practices
Model, Control, Verify, Recover
Record route
/capabilities/custom-technical-tools-clis
Content reviewed
2026-09-03

Established. Demonstrated by two or more published systems, so the pattern is repeated rather than singular.

Repeatable outcome

A tool that does one thing, states what it will do before doing it, and behaves the same way twice.

The recurring shape: inputs are validated at the boundary rather than trusted, the destructive operation is separated from the inspection that precedes it, the default is the safe mode, and the exit status means something a script can branch on. Failure modes are enumerated rather than discovered, which is what makes them testable — a tool whose errors are all “something went wrong” cannot be automated around.

These are operator tools. They are built to be run repeatedly by someone who needs the same answer each time, not to be demonstrated once.

Scope and boundaries

The evidence is three tools of different kinds: a document-maintenance utility, a read-only repository inspector, and a standards system with its own validation and drift gates. What they share is the contract discipline, not a language or a runtime.

This does not imply:

  • Availability of every tool. Two of the three evidence systems are published source; the third is private and says so on its own record. A capability being demonstrated is not the same as the demonstrating artefact being obtainable.
  • A general software-development service. This is a description of how tools of this shape get built here, not an engagement anyone can currently buy.
  • Support, maintenance, or a release commitment for anything built to this pattern.

Evidence systems

Published systems that demonstrate this capability, derived from those records rather than asserted here.

  1. L3D-SYS-003

    RepoLens

    A read-only repository intelligence tool that collects bounded evidence and publishes immutable snapshots without modifying what it inspects.

    Private while the first version is under review. It has not been released and carries no public promise.

  2. L3D-SYS-004

    Project Standards

    Versioned standards with schemas, adoption mechanisms, validation, and drift gates that a repository can be held to.

    Open source under Apache-2.0, with versioned releases. Read the source and adopt it directly.

  3. L3D-SYS-005

    docmend

    A document maintenance tool that separates scan, plan, apply, verify, and restore, defaults to a dry run, and can resume after interruption.

    Open source under MIT, with published releases. Install it or read the source.

  4. L3D-SYS-012

    Agent Pseudocode

    A pseudocode convention for instructing agents, with a validator, a formatter, and a language server so the convention is enforced rather than remembered.

    Open source under MIT. The source is readable and runnable; there are no published releases.

Active offerings

None. This capability describes work that has been done, and nothing on this site is for sale today. If that changes it will appear on the availability page as a record with its own scope and exclusions.

The capability index lists the rest. What can actually be obtained today is on the availability page, which is a different question from what has been demonstrated.