agent readiness
22 September 2026
Coding-agent performance assessment

docs-only

No coding agent has been measured on this repository yet.

Static readiness is 24 out of 100. The friction listed below is what a task suite would test.

Repository files only, no agent run · not a git repository · Static preflight only · prepared 22 September 2026

What we measuredThis repoWhat good looks likeReading
Finished and passed checksAttempts that produced a working change and passed every checkNot measuredGood: 80% or more. Poor: under 50%.Not measured
Needed a person to step inRuns where someone had to correct the agentNot measuredGood: 10% or less. Poor: over 30%.Not measured
Time to review one resultMedian engineer time to check what the agent producedNot measuredGood: 5 min or less. Poor: over 15 min.Not measured
Agent cost per finished changeAll agent spend, failures included, divided by changes that were acceptedNot measuredCompare with your engineer cost for the same change.Not measured
Do next
  1. Repository understanding. Capture non-obvious domain rules and point to authoritative implementations. Owner: tech lead of the repository · Done when: Ask the agent to trace a real business flow with file/symbol citations; have a maintainer verify it.
  2. Relevant-code discovery. Add concise module entry points where navigation failures are observed. Owner: tech lead of the repository · Done when: Give a representative issue without file hints; record correct localization and wrong-module detours.
  3. Instruction and context quality. Check guidance loading in the chosen agent; verify commands, remove contradictions and route task-specific detail. Owner: tech lead of the repository · Done when: Compare the same tasks with current versus corrected context under matched agent settings.

Then run a representative task suite: agree real tasks, acceptance criteria and one agent configuration, and measure.

01 / What the agent did

Performance on your work.

Each number shows what good looks like.
Definitions are under each card.

Finished and passed checks

Not measured

Good: 80% or more. Poor: under 50%.

Requires a task suite to measure.

Needed a person to step in

Not measured

Good: 10% or less. Poor: over 30%.

Requires a task suite to measure.

Time to review one result

Not measured

Good: 5 min or less. Poor: over 15 min.

Requires a task suite to measure.

Same result every time

Not measured

Good: 80% or more. Poor: under 50%.

Requires a task suite to measure.

Agent cost per finished change

Not measured

Compare with your engineer cost for the same change.

Requires a task suite to measure.

Time to a finished change

Not measured

Compare with your team's usual time for the same change.

Requires a task suite to measure.

The next step is a representative task suite.

Agree real tasks, acceptance criteria, an agent configuration and execution boundaries. Keep independent checks, human review and repeated attempts visible. No task commands were run by this static scan.

readylevel repo . --performance evidence.json --format all
02 / Where the codebase holds it back

Understand the friction.

What we saw in the repository.
Each item has a change and a retest.

CONTEXTobserved

Instruction and context quality

0 guidance and 0 context artifacts found. Presence does not establish actual loading, relevance or consistency. Referenced paths: 0 existing / 0 referenced.

Improvement & how to verify it

Action. Check guidance loading in the chosen agent; verify commands, remove contradictions and route task-specific detail.

Retest. Compare the same tasks with current versus corrected context under matched agent settings.

PLANobserved

Requirements and planning

0 planning artifacts found. PLAN.md is optional; acceptance criteria, correctness and freshness are not inferred from presence.

Improvement & how to verify it

Action. Make task scope, acceptance criteria, constraints and validation explicit in the team's existing task system.

Retest. Review whether the agent identifies required dependencies and asks about genuine ambiguity before changing code.

ISOLATEnot available

Change isolation

Language-aware dependency analysis is unavailable. File count and size do not measure semantic coupling.

Improvement & how to verify it

Action. Investigate boundary crossings and duplicated change sites using a real failure, not import count alone.

Retest. Make a bounded change and inspect unrelated edits, missed dependencies and regression outcomes.

7 areas a static scan cannot judge. Each has a suggested change and a retest.
UNDERSTANDnot measured

Repository understanding

0 architecture documents and 0 contract artifacts found. Business invariants and cross-file understanding remain untested.

Improvement & how to verify it

Action. Capture non-obvious domain rules and point to authoritative implementations.

Retest. Ask the agent to trace a real business flow with file/symbol citations; have a maintainer verify it.

DISCOVERnot measured

Relevant-code discovery

Static indexing does not establish whether an agent finds the right implementation and tests.

Improvement & how to verify it

Action. Add concise module entry points where navigation failures are observed.

Retest. Give a representative issue without file hints; record correct localization and wrong-module detours.

NAMINGnot measured

Naming and domain clarity

Semantic naming clarity and conflicting business meanings have not been assessed. Short identifiers are not automatically defects.

Improvement & how to verify it

Action. Review ambiguous domain terms, units and identifiers where concrete confusion is observed.

Retest. Test a task crossing the suspected terminology boundary; verify that meanings and units are preserved.

VERIFYnot measured

Verification ability

Test commands can be discovered statically; execution, oracle strength and the agent's verification choices are separate evidence.

Improvement & how to verify it

Action. Provide reproducible failures and independently reviewed acceptance/regression checks.

Retest. Require the fix to pass protected checks; verify that known incorrect solutions fail them.

CONTINUEnot measured

Session continuity

0 progress/handoff artifacts found. Successful recovery across sessions has not been tested.

Improvement & how to verify it

Action. Record remaining work, decisions and verification state in an appropriate durable artifact.

Retest. Resume in a fresh session; check for lost constraints, repeated work and premature completion claims.

EXTENDnot measured

Extension robustness

A static scan cannot establish whether an agent's first solution makes the next change harder.

Improvement & how to verify it

Action. Include follow-up requirements in the evaluation task set.

Retest. Extend the first implementation and check retained behaviour, rework and unnecessary duplication.

SAFEnot measured

Bounded operation

Repository-visible permission settings do not establish enforced sandboxing or safe runtime behaviour.

Improvement & how to verify it

Action. Agree isolated execution, data handling and permission boundaries before any task trial.

Retest. Use approved synthetic boundary tests without production credentials; inspect actual enforcement and escalation.

03 / Supporting engineering evidence

Engineering baseline.

24/100 static readiness

L0 · Unscored. 100% coverage of the versioned static checklist, not coverage of agent performance. Safety and verification remain gates, not offsettable bonus points.

Inspect static findings, levels and methodology
Decision viewThree overlapping, uncapped lenses over the same static evidence. They help prioritise remediation; they do not predict task success.
FOUNDATIONS8

Engineering foundations

Reproducibility, verification, structure, and repository knowledge.

2/4 dimensions scored
CHANGE_SAFETY43

Change safety

Controls that help a change remain reviewable, testable, and bounded.

3/4 dimensions scored
AGENT_OPERABILITY33

Agent operability

Static preconditions for supervised coding-agent work.

3/4 dimensions scored
100%evidence coverage

7 stable checks scored · 16 not applicable · 0 unavailable. Coverage reports what could be resolved, not how healthy the repository is.

ObservedRepository evidence

Files, configuration, and available git history.

Not runCommand execution

Setup, build, tests, lint, and type checks.

Separate evidenceAgent task evaluation

See the performance section. Static checks do not establish task success.

Not assessedProduction delivery

Deploy, rollback, reliability, and operational controls.

Repository profile. documentation (operator supplied confidence). Portfolio manifest classification; verify with the repository owner.

Knowledge inventory (presence only). Agent guidance: none found. Planning: none found. Documentation: 2 files. Missing PLAN.md is not a failed control; file presence does not establish quality or freshness.

Structure diagnostic. Language-aware dependency analysis not available for this repository; no semantic coupling or naming claim is made.

Decision boundary. Use as a portfolio preflight and remediation baseline, not as proof of autonomous delivery.

Where to start
  1. Add a workflow that runs the test command on pull_request and push.

  2. Add `make setup` (or the ecosystem equivalent) that works from a clean clone, and name it in AGENTS.md.

  3. Add .devcontainer/devcontainer.json with postCreateCommand running your setup command.

Next level L1 Discoverable. Static evidence identifies the repository's setup, test, and usage entry points; those commands were not executed. Requires ENV-02 >= 5 (is 0).

Evidence dimensionsRisk-weighted static signals. Bar length is score; the small figure is weight.
Reproducibility
010
Verificationno conclusive checksn/a10
Delivery controls
07.5
Security boundaries
107.5
Changeabilityno conclusive checksn/a7.5
Agent guidance
05
System knowledge
42.5
Work definition
02.5
Agent controlsno conclusive checksn/a2.5

pass partial fail experimental, not counted not applicable

OpportunitiesEvery counted check below ten, in order of what it is worth. Points are the change to the score if the check reached ten.
01

CI runs tests on pull requests

+22
CI-01·0 of 10·No CI configuration found.

Add a workflow that runs the test command on pull_request and push.

Why it matters DORA's AI capabilities model names strong version control and fast feedback as the amplifiers of AI gains; without CI an agent's PR is unverified. [1]

02

One-command setup

+15
ENV-02·0 of 10·No way to set up the project without a person.

Add `make setup` (or the ecosystem equivalent) that works from a clean clone, and name it in AGENTS.md.

Why it matters Every agent vendor's guidance opens with "install, run, test commands"; Codex's own docs say many quality issues are really setup issues. [1][2]

03

Container or devcontainer definition

+15
ENV-03·0 of 10·No container definition.

Add .devcontainer/devcontainer.json with postCreateCommand running your setup command.

Why it matters Devcontainers are what Copilot, Cursor cloud, and Claude Code cloud consume; EnvBench shows agents build under 30% of environments without one. [1][2]

04

Cross-tool instruction file present

+15
INSTR-01·0 of 10·No agent instruction file.

Add a root AGENTS.md with commands, layout, conventions, and completion criteria. Bridge CLAUDE.md to it (INSTR-02).

Why it matters Two controlled studies found instruction files do not change pass rates, which is why this dimension is Medium; they do change cost and every vendor reads them. [1][2][3]

05

Issue templates require reproduction and expected behaviour

+7
GIT-01·0 of 10·No issue templates.

Convert the bug template to a form with required Steps to reproduce, Expected, Actual, and Version fields.

Why it matters Underspecified issues were 38% of the flaws in the original SWE-bench; human-written requirements tripled agent success in a controlled study. [1][2]

06

README covers setup, usage, and testing

+5
DOCS-01·4 of 10·README exists but lacks operational sections or commands.

Add Getting started, Usage, and Development sections with one command each.

  • README.md

Why it matters Copilot's coding agent reads the README before coding; documentation reduces API hallucination when it is specific. [1][2]

DiagnosticsExperimental checks. Computed and shown; not counted.
CI-06Git hooks manager configured0No git hooks manager.
DOCS-03Architecture decisions written down0No architecture document or decision records.
DOCS-04CONTRIBUTING present0No CONTRIBUTING file.
ENV-05Runtime version pinned0No runtime version pin (.nvmrc, .python-version, .tool-versions, rust-toolchain, go directive, engines).
GIT-02PR template asks for verification0No PR template.
GIT-03CODEOWNERS present0No CODEOWNERS.
STRUCT-08No large binaries in the tree10No tracked files over 5 MiB outside asset directories.
Passed
1 checks at ten
SAFE-01No committed secretsNo high-confidence secret patterns and no tracked .env with values.
Not applicable
30 checks did not apply or had no data
CI-02Linter configured for the primary languagenot applicable to this repository
CI-03Formatter configurednot applicable to this repository
CI-04Type checking configurednot applicable to this repository
CI-07Lint or format enforced in CInot applicable to this repository
DOCS-02API contracts in the repositorynot applicable to this repository
ENV-01Every manifest has a tracked lockfilenot applicable to this repository
ENV-04Agent cloud setup file present and validnot applicable to this repository
GIT-04Historical change size is smalldata not available:
GIT-05Repository is maintaineddata not available: not a git repository
GIT-06Conventional commit messagesdata not available:
INSTR-02Claude and AGENTS.md instructions do not divergenot applicable to this repository
INSTR-03Instruction files within vendor size bandsnot applicable to this repository
INSTR-04Instruction files state build, test, and lint commandsnot applicable to this repository
INSTR-05Referenced paths existnot applicable to this repository
INSTR-07No directory-tree dumps in instruction filesnot applicable to this repository
INSTR-10Local-only agent files are untrackeddata not available:
SAFE-02Secrets excluded from agent readsnot applicable to this repository
SAFE-04Destructive commands guardednot applicable to this repository
SAFE-05Sandbox configurednot applicable to this repository
STRUCT-01Share of source files over 1,000 linesnot applicable to this repository
STRUCT-02Repository size tiernot applicable to this repository
TOOL-01Committed agent settings are validnot applicable to this repository
TOOL-02Auto-format or lint hooknot applicable to this repository
TOOL-03Verification hooknot applicable to this repository
TOOL-04MCP configuration valid and scopednot applicable to this repository
TOOL-05Skills and subagents well-formednot applicable to this repository
TOOL-07Build artefacts ignorednot applicable to this repository
VERIFY-01Test framework configured with test filesnot applicable to this repository
VERIFY-02One documented test commandnot applicable to this repository
VERIFY-03Test-to-source file rationot applicable to this repository
Method
Score
The risk-weighted mean of nine static repository dimensions, 0 to 100. It is a prioritisation index, not a probability that an agent will succeed.
Level
An evidence band earned only when every static check in its gate passes. It describes repository preconditions. It does not certify command execution, autonomous task completion, merging, or production delivery.
Coverage
The share of stable checks resolved from available evidence. Not-applicable checks are resolved; unavailable or errored checks reduce coverage. Coverage is shown separately so missing evidence cannot hide behind the score.
Points
What the score becomes if one check is fixed and nothing else changes. The opportunities are already in that order.
Determinism
The static scorer uses no model judgment. Agent runs can vary; imported judgments are separate evidence. Two static runs with the same inputs hash saw the same indexed repository.
inputs sha256:d3949061beb0dae00275e13a18f9f7ee0f2c2edf6f8edaa4fcd0ba2d03028cad
rubric sha256:add0515531acf415c6a9dd5e7f9e1e7585f47c682087c8055ed58c3fabe3990f
Not in static score
Command execution, coding-agent task success, semantic coupling, naming quality, deployment, rollback, reliability, and production operations. Any execution or language-aware diagnostic above is reported separately and cannot change this score.
Benchmark
A percentile against a validated corpus is not yet published. This report shows none rather than inventing one.
04 / Evidence you can inspect

Scope, provenance & limits.

Agent performance has not been measured. This scan identifies observable codebase friction and the tasks needed to test it; it does not predict success.

How to read the numbers. Every figure comes from recorded runs of one agent configuration on one repository revision. Missing evidence is shown as not measured, never estimated. "What good looks like" bands are decision thresholds set by rubric apr/0.1.0, not industry benchmarks or percentiles; a band never authorises unattended merging, and anything that broke existing behaviour or a safety rule overrides every band.

Only repository files and supported static diagnostics were assessed. Task completion, human effort, cost, session recovery and production performance are unmeasured.

No universal autonomy certification, validated percentile or calibrated success prediction. Small and selected task sets limit generalisation. Missing evidence stays unknown. Imported reviewer and independence claims are supplied evidence, not attested facts. Production delivery is not established.