[GH-ISSUE #254] Skip binary-tool setup for source-repository analysis #149

New issue

Closed

opened 2026-10-15 18:21:39 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-15 18:21:39 +00:00

Owner

Originally created by @morluto on GitHub (Jul 15, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/254

Problem

The rea-analysis skill triggers broadly for requests to reverse engineer or understand an application. Its workflow then requires npx -y rea-agents doctor before distinguishing between:

binary, package, artifact, or runtime analysis, where REA tooling is relevant; and
architecture analysis of a repository whose complete source code is already available, where normal source inspection is sufficient.

In practice, a request to inspect a Rust repository and compare its architecture caused the agent to start the REA readiness workflow. This looked like an unnecessary installation/setup path and distracted from the source-level task.

Suggested behavior

Add an early routing decision before Ensure REA is ready:

Determine whether complete source code is available and whether the requested evidence requires binary/runtime inspection.
For source-only architecture analysis, use normal repository tools and skip REA readiness/setup.
Run rea-agents doctor only when binary, package, artifact, decompilation, or runtime analysis is actually needed.

The skill trigger description could also distinguish compiled-application reverse engineering from ordinary source-code investigation.

Why this matters

Avoids unnecessary network/setup commands.
Prevents users from thinking binary tooling is required for ordinary repository analysis.
Keeps REA's evidence workflow focused on cases where its specialized tools provide additional authority.

No installation completed in the observed run because the readiness command was interrupted.

Originally created by @morluto on GitHub (Jul 15, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/254 ## Problem The `rea-analysis` skill triggers broadly for requests to reverse engineer or understand an application. Its workflow then requires `npx -y rea-agents doctor` before distinguishing between: - binary, package, artifact, or runtime analysis, where REA tooling is relevant; and - architecture analysis of a repository whose complete source code is already available, where normal source inspection is sufficient. In practice, a request to inspect a Rust repository and compare its architecture caused the agent to start the REA readiness workflow. This looked like an unnecessary installation/setup path and distracted from the source-level task. ## Suggested behavior Add an early routing decision before `Ensure REA is ready`: 1. Determine whether complete source code is available and whether the requested evidence requires binary/runtime inspection. 2. For source-only architecture analysis, use normal repository tools and skip REA readiness/setup. 3. Run `rea-agents doctor` only when binary, package, artifact, decompilation, or runtime analysis is actually needed. The skill trigger description could also distinguish compiled-application reverse engineering from ordinary source-code investigation. ## Why this matters - Avoids unnecessary network/setup commands. - Prevents users from thinking binary tooling is required for ordinary repository analysis. - Keeps REA's evidence workflow focused on cases where its specialized tools provide additional authority. No installation completed in the observed run because the readiness command was interrupted.

bsakaguchi

2026-10-15 18:21:39 +00:00

closed this issue
added the
area: setup

enhancement
labels

bsakaguchi commented

2026-10-15 18:21:41 +00:00

Author

Owner

@morluto commented on GitHub (Jul 17, 2026):

Implemented on the current working branch.

The canonical reverse-engineer-anything skill now makes an explicit request-routing decision before readiness: complete-source architecture/code-flow analysis uses ordinary repository tools and skips REA doctor/setup, while binary, package, built-artifact, decompilation, passive-runtime, replay, or behavior-dependent claims continue into the REA workflow. The skill version is bumped to 22 so managed installations upgrade transactionally, and tests/setupSkill.test.ts asserts the installed routing contract.

Validation run:

npx vitest run tests/setupSkill.test.ts --coverage.enabled=false
npm run format:check
npm run docs:check

Generated package metadata and the product catalog were refreshed.

@morluto commented on GitHub (Jul 17, 2026): Implemented on the current working branch. The canonical `reverse-engineer-anything` skill now makes an explicit request-routing decision before readiness: complete-source architecture/code-flow analysis uses ordinary repository tools and skips REA doctor/setup, while binary, package, built-artifact, decompilation, passive-runtime, replay, or behavior-dependent claims continue into the REA workflow. The skill version is bumped to 22 so managed installations upgrade transactionally, and `tests/setupSkill.test.ts` asserts the installed routing contract. Validation run: - `npx vitest run tests/setupSkill.test.ts --coverage.enabled=false` - `npm run format:check` - `npm run docs:check` Generated package metadata and the product catalog were refreshed.

bsakaguchi referenced this issue

2026-10-16 21:12:01 +00:00

[PR #219] [CLOSED] fix(process): restrict cleanup to captured group leaders #1872

bsakaguchi referenced this issue

2026-10-16 21:12:07 +00:00

[PR #230] [MERGED] fix(process): gate sampling on initialized PTY roots #1881

bsakaguchi referenced this issue

2026-10-16 21:12:09 +00:00

[PR #233] [MERGED] fix(process): settle exited zombie groups #1883

bsakaguchi referenced this issue

2026-10-16 21:12:11 +00:00

[PR #235] [MERGED] fix: address setup, upgrade, and process test regressions #1885