hamsa software
← Hamsa Software

FOR ENGINEERING TEAMS / A SEPARATE SPECIALIST ENGAGEMENT

Know what your
green checks mean.

An engineering review of the access and verification around one coding-agent workflow. Built for teams preparing to give agents more responsibility.

Discuss your workflow See a sample finding

Delivered by Hamsa Jama · London & remote · scope and fixed fee agreed before work

Looking to improve payment and finance operations? Explore the Payment Operations Pilot.

WHEN THIS IS USEFUL

Your agent can do more.
Your checks need to keep up.

A BOUNDED STARTING POINT

One workflow, examined properly.

IN SCOPE

One agreed path.

  • One repository and coding-agent setup
  • One CI and release approval path
  • Permissions, credentials and editable verification inputs
  • Agreed scenarios in an isolated copy
YOU RECEIVE

Evidence and decisions.

  • Workflow and access map
  • Reproducible findings with stated limits
  • Prioritised changes and a walkthrough
  • Implemented improvements and repeat checks where included in scope
BEFORE WE START

A written agreement.

  • Scope, fixed fee and delivery dates
  • Access and confidentiality arrangements
  • Your team’s involvement and named owners
  • Handover, support and ongoing costs

The initial review covers the selected workflow. Wider penetration testing, compliance certification and guarantees about all future agent behaviour are outside this offer.

AN EXECUTED EXAMPLE

A passing run.
An overlooked regression.

The same faulty implementation is tested against two different sets of checks. The result changes when the original test is restored.

Original code · original checks2/2 pass
Faulty change · weakened checks1/1 pass
Same faulty change · original checks1/2 pass · 1 fail

Reproduce it.

The script creates temporary files, runs Node’s test runner for all three cases, checks the expected outcomes and removes the temporary workspace.

node scripts/verification-example.mjs

Download the runnable script · Read the recorded results

What this establishes.

This is a deterministic, scripted fixture. No AI model, coding agent or Tamperward runtime is used. It demonstrates why a changed check can hide a regression; it does not establish that an agent will do this or that any particular system is secure.

SAMPLE FINDING / FROM THE FIXTURE ABOVE

Inspect the kind of
evidence you receive.

An illustrative report entry based on the executed example. Customer findings depend on the agreed workflow and observed results.

HS / VERIFICATION REVIEWSPECIMEN · 01

A removed test hides a quantity regression.

Observed
Faulty code passes the remaining single-item test. Original checks fail the multi-item case.
Consequence
A green run can approve code that calculates the wrong total for multiple items.
Recommended change
Protect the baseline checks and review changes to tests and CI separately from the proposed implementation.
Repeat check
Restore the original checks and confirm that the regression is detected. Verify the chosen controls in the actual workflow.
Limit
Two tests, one scripted regression. No claim of complete coverage or deployment certification.

DIRECT ACCESS TO THE ENGINEER

Work with Hamsa.

Hamsa Jama scopes and delivers the engagement. His professional background includes payment and subscription systems at Which?, finance automation at Cazoo, and document processing at Tata Consultancy Services.

Your scope names the delivery dates, required access and time from your team. Handover includes the agreed findings, reproduction steps and a walkthrough. Any ongoing support or additional specialists are agreed explicitly.

Read the professional experience

Bring one workflow.
Start with a conversation.

Discuss a coding-agent review

Want your engineering team to practise the approach? Explore the supported coding-agent workflow workshop.