Sruja harness

Skill sruja-ai/sruja/skills/sruja-harness

A CLI-first autonomous coding agent where the actor never grades itself. Every edit is grounded in your repo's real topology and verified by an independent deterministic grader (drift, lint, verify-task, intent) before it ships. Closed-loop: comprehend → plan → execute → critique → replan. Also a passive MCP harness.

Install

npx -y skills add sruja-ai/sruja --skill sruja-harness

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

24 stars24 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Verification harness for any AI agent skill. Use with any other skills; before marking a task done, run verify-task. On failure, record a correction learning. Ensures lint, tests, drift, and intent checks pass regardless of which skill generated the code.

The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.3 KB, as published. Nobody here has run it

Sruja Harness Skill

Universal verification adapter — works with any coding, bugfix, or review skill.

Quick Start

Install alongside other skills:

npx skills add https://github.com/sruja-ai/sruja --skill sruja-harness
npx skills add https://github.com/sruja-ai/sruja --skill sruja-architecture

Use with any skill: This skill does not generate code. It verifies output from other skills.

Before marking a task done:

sruja verify-task --profile coding -r .

On verify failure:

sruja agent record -r . \
  -c "<task context>" \
  -H "<what was tried>" \
  -o failed \
  -g "<correction for next time>"

Rules

Always verify before done — Run sruja verify-task with the appropriate profile before declaring a task complete.
Profile selection:
- coding — New features, refactors (default)
- bugfix — Bug fixes (use --file to target the bug location)
- review — Pre-merge review of changes
- arch — Architecture/DSL changes
On failure: Record a correction learning before retrying.
On success: Optionally record an affirmation learning ("what worked").

Integration with Other Skills

Coding Skill (e.g. /build) → sruja-harness verify → record_learning → done

This skill does not replace coding skills — it ensures their output passes Sruja's architecture gates.

Profiles

Profile	Verifies
`coding`	`lint repo.sruja` + `just check` (or `make check`) + `drift`
`bugfix`	`focus --file` + `just check` (or `make check`) + `intent check`
`review`	`review -f json` + `intent check` + `drift`
`arch`	`lint repo.sruja` + `drift` + `intent check` + `review`

References

docs/HOST_AGENT_INTEGRATION.md — Full integration contract
docs/plans/AGENT_DELIVERY_PLAN.md — Delivery roadmap