Ai C Plan
ReviewAudited by ClawScan on May 10, 2026.
Overview
The skill is coherent with its autopilot coding purpose, but it asks the agent to change files, run commands, apply migrations, and commit automatically with limited user checkpoints.
Review the plan and architecture files before use, run only on a safe development branch, confirm the target workspace path, keep backups, and ensure any database configuration points to a local or disposable development database before allowing autopilot execution.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent could make many codebase and database changes before the user has a chance to review them.
The instructions authorize automatic file changes, database/schema migrations, and broad git staging/commits without normal per-step approval.
Bu skill, kullanıcıdan her adımda onay istemeden ilerler. ... Database/schema değişikliği yapılıyorsa migration üret ve uygula. ... git add .
Use only in a trusted development repository, preferably on a separate branch with backups, and require explicit approval before migrations, broad git staging, or commits.
If invoked from a different workspace, the agent may still read or write files under the hard-coded AI-C path.
The skill gives a workspace-only rule but also requires writing a hard-coded absolute progress path, which can conflict with the intended workspace boundary.
Workspace dışına yazma. ... Progress kaydı tek kaynak olarak `/home/adige/AI-C/progress.json` üzerinde tutulur.
Replace hard-coded absolute paths with user-confirmed workspace-relative paths or require the user to confirm the target project directory before execution.
Repository scripts or migrations may execute local code and alter local services or databases automatically.
Running repository lint/test/build scripts and applying migrations is expected for a development skill, but here it is paired with autopilot execution and limited user checkpoints.
Komutlar: lint/test/build (repo yapısına göre) ... Prisma migration (`apps/api` içinde) + migration dosyaları
Run in an isolated development environment, verify scripts first, and ensure database environment variables point only to disposable or local development databases.
A single incorrect plan or architecture instruction could lead to a chain of unwanted changes throughout the project.
The skill automatically expands the plan into dependent steps and continues through runnable work items, so a bad plan entry or mistaken dependency can propagate across many files and commits.
Planı oku → tüm ST-* satırlarını çıkar. Bir DAG oluştur ... Kullanıcıdan onay istemeden sıradaki runnable maddeye geç.
Review and trust the plan files before starting, limit the number of steps per run, and inspect changes between batches.
Untrusted or stale local guidance could influence future automated steps.
The skill relies on local prompt-like architecture content and a persistent progress file to guide future actions; this is purpose-aligned but should only be done with trusted project files.
docs/architecture/AI_CONSULTANT_SYSTEM_PROMPT.md dosyasını oku ... Progress kaydı tek kaynak olarak `/home/adige/AI-C/progress.json` üzerinde tutulur.
Inspect the architecture prompt, plan files, and progress.json before running or resuming the skill.