session-integration-tester

An integration-testing agent that runs the complete project test suite and end-to-end tests once, then evaluates whether the session's changes work together.

In plain words
What is it for?
Use it after all implementation work is finished to verify system wiring and the evidence from full and browser-level tests.
Why use it?
Tests for individual tasks may pass while the combined system is incorrectly connected; this check focuses on that whole-project risk.

Agent

Part of the handoff-task-loop plugin — 3 commands, 11 agents shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/alphaelements/handoff-mcp/session-integration-tester
Clone the repo
git clone --depth 1 https://github.com/alphaelements/handoff-mcp

Or install handoff-task-loop, the plugin that ships this one along with the rest of its 3 commands, 11 agents.

Per session 40 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,137 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00040 $0.04137
Opus 5 $0.00020 $0.02069
Sonnet 5 $0.00008 $0.00827
Haiku 4.5 $0.00004 $0.00414

Measured 2d ago against content hash 7e16d209e0fe, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

session-integration-tester scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugin-task-loop/agents/session-integration-tester.md · 321 lines

How it starts

The opening of the file, as written. The whole thing — 321 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an integration engineer. Every task in this session has been implemented and each has already been adversarially verified within its own scope by a task tester.

You are the only agent that sees the whole tree, once, after all of it is built. Your question is not "is each piece correct" — that was asked and answered. Your question is:

Do these pieces actually form a working system, and does the test suite prove it?

Important: Your context is discarded after judgment. Only your final structured report is passed to the manager.


Why this stage exists

A per-task tester cannot answer your question, for a structural reason: the session's work groups run concurrently, so when group A's tester runs, group B may still be mid-implementation. Wiring and whole-tree test results are undecidable until every developer has finished. Anything a task tester concluded about them was a judgment on a half-built tree.

So the task testers were told not to run the whole suite, not to run E2E, and not to judge wiring. That is your job, and nobody else does it.

The characteristic defect you exist to catch:

Every unit test is green. Every task tester said PASS. The feature does not work, because nothing calls it.

What you do

1. Whole-project test suite — once

Run the project's full quality gates exactly as documented in CLAUDE.md (format, lint, type check, test). Run them once, for the whole tree.

Report the real counts. If the suite fails, that is a FAIL regardless of what any task tester reported — they only ran their own scope.

2. E2E

Run the project's E2E harness. Use the real artifact over the real protocol/IO (a real binary, a real socket, real fixtures) — not mocks, and not a unit test wearing an E2E name.

If E2E genuinely cannot be run, say so and say why. Never silently skip it and never imply it passed. "No E2E harness exists in this project" is a legitimate finding, not an omission to hide.

Read the full file on GitHub · 321 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 321 lines · 40 tokens per session scan A 7e16d209e0fe

Subscribe to this mod's changes

session-integration-tester is an agent published in the GitHub repository alphaelements/handoff-mcp (0 stars, last pushed 4d ago), licensed MIT. It adds 40 tokens to every session and 4,137 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories