Security audit
Security checks for vulnerabilities and agentic risk
This is a startup-coaching skill that creates and resumes local business-planning documents, with scoped and disclosed persistence but no hidden code, network access, credential use, or destructive behavior.
Before installing, expect the skill to save business model, execution, and brand documents in ~/opc-guide/ and to read them later when resuming. Avoid including sensitive customer names, pricing, or strategy details unless you are comfortable storing them locally in those Markdown files.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
- **Have revenue, want people to remember you** → Start with `brand.md`
- **Full journey** → Phase 1 → 2 → 3 in order
### Document Output Rules
After each Phase, write the output document to the user's local directory:
The description lists phrases like "I have an idea," "should I do this," and "how do I execute," which are common, non-domain-specific expressions that could match many unrelated conversations. The file does not provide exclusion conditions or narrow context constraints to prevent unintended invocation.
Phrases such as "stuck," "differentiation," and "don't know where to start" are vague and can apply to many domains beyond one-person-company coaching. The table maps them to phases but does not define contextual constraints or exclusions, increasing the risk of accidental routing.
The skill instructs the agent to write persistent files to a local directory and later read them automatically on phrases like 'continue from last time' without any explicit user consent, transparency, or scoping controls. This creates a privacy and persistence risk because business plans, brand strategy, and potentially sensitive commercial information may be stored and reloaded from the local filesystem unexpectedly.
The skill instructs writing user-supplied business information to a persistent local file path without warning the user that their responses may be stored on disk. Because the collected content includes names, employers, pricing, customers, and business plans, this can create unintended retention of sensitive information and possible later disclosure to other tools, users, or processes on the host.
The skill instructs the agent to write a file under the user's home directory without any safety notice, confirmation step, or constraints to prevent unintended overwrites. Even though the target path is scoped to ~/opc-guide/, modifying local files is a side effect that can overwrite prior work or create persistence the user did not explicitly approve at execution time.
The red-flag example includes the phrase "feature堆砌," which introduces a Chinese-language term into an otherwise English-only skill. Because the skill does not give the user any language preference or explain multilingual output, this creates a language/locale policy inconsistency.
Line L074 contains Chinese text ("认同") embedded in otherwise English instructions. This introduces a language/locale constraint in the skill content without user opt-in or any explanation that multilingual output is intended or required.
The output format requires a field labeled 衡量标准, introducing Chinese text into the generated output without offering the user a language choice. This can violate language/locale policy when the rest of the skill is in English and no opt-in or justification is provided.