Use the Apiiro MCP tool `apiiro_secure_prompt` when Apiiro MCP is configured in the client and the user wants security guidance added to a coding task before implementation. Trigger on secure-prompt via MCP, or when the user asks for security-aware prompt enhancement before coding (auth, new APIs, persistence, secrets, integrations).
How this skill is triggered — by the user, by Claude, or both
Slash command
/apiiro-guardian-agent:apiiro-mcp-secure-promptThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Enhance a coding prompt with security context through the **Apiiro MCP** tools. Relevant whenever you have **Apiiro MCP configured** in your environment (IDE, agent, or other MCP client).
apiiro_secure_prompt)Enhance a coding prompt with security context through the Apiiro MCP tools. Relevant whenever you have Apiiro MCP configured in your environment (IDE, agent, or other MCP client).
| Requirement | Notes |
|---|---|
| Apiiro MCP configured | MCP client is connected to an Apiiro MCP server; tools are available after a successful session. |
| Tool available | If apiiro_secure_prompt does not appear, the server may not expose that tool—check with your Apiiro administrator or deployment docs. |
| Auth | Follow whatever authentication your organization configured for that MCP server. |
apiiro_resolve_repository_key — Pass repository_name (optional branch, remote_url). Returns a repository_key string or an error payload.apiiro_secure_prompt — Pass the original user task as developer_prompt, the repository_key from step 1, and optionally file_path (current file for scoped context).If you already have a valid repository_key for the repo, skip step 1.
Resolve key with less ambiguity: When the repo might not match by name alone, pass remote_url from git remote get-url origin and the current branch from git branch --show-current.
Run in parallel with exploration: Do not block codebase reads or searches on the MCP call—invoke apiiro_secure_prompt (after you have or can cheaply obtain repository_key) alongside looking up relevant files, unless the task strictly depends on the enhanced text before any read.
| Argument | Required | Purpose |
|---|---|---|
developer_prompt | Yes | Verbatim coding task or feature request to enrich. |
repository_key | Yes | From apiiro_resolve_repository_key, or an existing key for that repository. |
file_path | No | File path string for tighter scope (e.g. the file being edited). |
Return value: On success, the enhanced prompt text (plain string). On failure, a structured error string from the MCP server.
| Situation | What to do |
|---|---|
| Task is one line or vague (e.g. “check auth flow”) | Expand developer_prompt to a short spec: goal, constraints, environment; add file_path to the main file under review. |
| Output looks unchanged or nearly identical to input | Retry once with file_path and a fuller developer_prompt; if still thin, proceed and apply normal secure-design practices (input validation, least privilege, secrets handling, audit logging) per org standards. |
| Monorepo or fork | Prefer accurate repository_name + remote_url so apiiro_resolve_repository_key maps to the right Apiiro project. |
Some environments document a fixed repository_key for a specific repo (e.g. to skip resolve). Use only if your team’s docs list one; otherwise always resolve via apiiro_resolve_repository_key.
npx claudepluginhub apiiro/mcp-serverBlocks Edit/Write/Bash actions until Claude investigates importers, data schemas, and user instructions. Improves output quality by forcing concrete facts before edits.