Security audit
Security checks for vulnerabilities and agentic risk
This skill sends selected coding-work summaries to Perfly, with clear limits on what can be inspected and submitted.
Install this only if you intend to use Perfly Agent Sync and are comfortable sending concise worklog metadata to Perfly. Avoid including confidential branch names, issue keys, or commit titles if those identifiers reveal sensitive project information.
The text states that this package contains the 'English canonical skill instructions,' which imposes a specific language in the skill description. Under the policy, forcing a specific language without offering a user language choice or documenting a justified locale constraint is a natural-language policy violation.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Configure the hosted Perfly MCP endpoint in the user's MCP client:
https://api.perfly.dev/mcp
OAuth discovery is the recommended setup. Manual bearer tokens are fallback-only for headless, SSH, devcontainer, CI, or remote sandbox workflows that cannot complete browser OAuth.
The skill instructs the agent to transmit user-derived worklog metadata to an external hosted endpoint, which is a real data exfiltration boundary even if the destination is the intended Perfly service. The risk is moderated by explicit privacy limits, preview-before-write, and a ban on uploading source code or secrets, but the skill still enables outbound transfer of potentially sensitive project metadata such as branch names, commit titles, and issue keys.
Production MCP endpoint:
https://api.perfly.dev/mcp
Prefer MCP OAuth discovery when the client supports browser authorization. The user should only need to add the hosted MCP endpoint.
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.
If the user asks to include sensitive material, summarize at a higher level and explain that Perfly Agent Sync accepts summary metadata only.
If MCP is unavailable, explain how to set up Perfly Agent Sync with the hosted MCP endpoint first. Do not ask the user to paste an ingestion token directly into chat; if OAuth cannot work in the user's environment, direct them to create a fallback token in the Perfly app setup page.
The manifest description is broad enough to match generic requests about coding summaries or worklogs, which can cause the skill to activate in contexts beyond its intended scope. In this skill, overbroad activation is notable because the declared tools can submit work items to an external Perfly service, increasing the chance of unintended data sharing or user confusion about when outbound syncing may occur.
No suspicious patterns detected.