Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed workflow-supervision skill that may create a local project checkpoint file, but it does not include hidden code, network access, credential handling, or destructive instructions.
Install this only if you want an agent to impose a structured completion and validation loop on complex work. Expect it to read project rules and maintain a local PROJECT-CHECKPOINT.md for larger tasks; review or ignore that file if your repository has strict tracking, privacy, or workflow rules.
The skill advertises itself for extremely broad, common task phrases like complete, fix, continue, validate, and resume, which creates ambiguous invocation boundaries. In agent systems, overbroad activation can cause this supervisory skill to attach to many unrelated tasks, unexpectedly altering execution flow, file writes, and persistence behavior across projects.
The README presents all user-facing instructions and behavior descriptions only in Chinese. Under the policy for natural-language violations, forcing a specific language without giving users a choice or documenting a justified locale constraint is reportable.
The compatibility, metadata description, and argument hint are written only in Chinese, and later trigger examples also emphasize Chinese phrases alongside English ones. This introduces a language/locale bias in the skill's natural-language interface without stating that users may choose their preferred language or that the skill is region-specific.
The guidance to use the skill whenever ANY broad project-like condition applies makes the dispatcher highly permissive and subjective. That increases the chance of unnecessary autonomous supervision, repeated looping, and state-file creation in contexts where the user did not intend persistent workflow control.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
**Inspect** — before acting: current goal, the user's latest message, project
`AGENTS.md` / `QWEN.md` / spec / README, `PROJECT-CHECKPOINT.md` if present,
current file state, prior results, the last validation outcome. Never restart
from scratch without checking; never redo already-VERIFIED work.
**Select** — pick the highest-priority unresolved ledger item (P0 first).
The directive for automatic maintenance that says to never ask the user to maintain the checkpoint, combined with instructions to create and update PROJECT-CHECKPOINT.md automatically, authorizes persistent file writes without explicit user confirmation. In shared or sensitive repositories, this can create unintended state changes, leak task context into tracked files, or interfere with project workflows.
agents or deliverables, has complex acceptance, will be iterated later, or where
a dropped requirement is expensive.
**Automatic maintenance** (never ask the user to maintain it):
- First complex task in a project with no checkpoint: create it, extract goal,
build the ledger, fill Definition of Done, constraints, `NEXT ACTION`.
- At meaningful phase boundaries update `DONE` / `IN_PROGRESS` / `TODO` /
No suspicious patterns detected.