Security audit
Security checks for vulnerabilities and agentic risk
This is a game-architecture reference skill with an optional document-writing workflow, and I found no hidden code, credential access, network execution, or unsafe persistence.
Installers should expect this skill to load game-architecture reference text and, when used in standalone mode, create or update planning files under architect/. Review existing architect/ files before asking it to refactor them if you have manual edits you want preserved.
The skill description is broadly scoped to trigger on essentially any game-system design or planning request across many domains, which increases the chance of unintended activation. Over-broad activation can inject irrelevant or excessive instructions into unrelated conversations, causing context pollution, reduced agent reliability, and making prompt-routing easier to manipulate.
Output size or generation rate is not bounded. Unbounded output enables denial-of-service through resource exhaustion, log flooding, or context-window stuffing.
### Traversal Strategies
* **Depth-First Search (DFS):** Explores as far as possible along each branch before backtracking. Useful for maze generation.
* **Breadth-First Search (BFS):** Explores neighbor nodes first. Guarantee shortest path in unweighted graphs. Useful for flood-fill logic (e.g., "area of effect" calculation).
### Precalculation & Caching
* **Results Save To File:** Bake static paths or navigation data (NavMesh) during build time.
The workflow explicitly directs the agent to create an architect/ directory and write multiple files into the user's workspace, but it does not require prior confirmation or warn that local files will be created or modified. In an agent setting, silent workspace writes can surprise users, overwrite existing artifacts, or be chained with other behaviors to cause unintended state changes.
The refactor extension instructs updating an existing output file in place after reading it, but it does not require a backup, diff preview, or explicit overwrite confirmation. This can lead to accidental loss of user edits or destructive modification of prior work, especially because the file may have been manually changed by the user.
No suspicious patterns detected.