How this skill is triggered — by the user, by Claude, or both
Slash command
/zeropath:findingsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Read-path entrypoint. Inline, no subagent - this is just an MCP call
Read-path entrypoint. Inline, no subagent - this is just an MCP call and a render.
Get the repo binding once. One bash call:
node "${CLAUDE_PLUGIN_ROOT}/scripts/current-binding.mjs"
Branch on the result:
ok: true → use repositoryIds: ["<repositoryId>"] and
organizationId: "<orgId>" in the MCP call. Mention the repo in
your reply ("Open findings in <orgName>/<repoUrl>:").ok: false, reason: "resolve_failed" → ask in plain chat
whether to add the repo to ZeroPath via repositories.addByUrl
or list org-wide for now. Don't auto-add.Don't probe creds, the CLI, or config in chat - the auto-bootstrap handles all that.
Call issues.list with sensible defaults:
{
"offset": 0,
"limit": 25,
"statuses": ["PENDING_REVIEW"],
"sortBy": "score",
"sortOrder": "desc"
}
Layer on scoreLevels, vulnerabilityClasses,
repositoryBranches, searchQuery based on the user's filter.
Render compactly. One row per finding:
id | severity | vuln class | repo | file:line | title.
Cap output to ~20 rows; surface counts.total and offer to page.
Recognize adjacent intents and route to other MCP tools. Same
underlying server exposes more than issues.list. Reach for these
when the user's intent points at them:
lodash?" → sca.list.vulnerabilities.search.endpoints.search with the
bound repositoryId.stats.summary
(+ stats.assets).code.search with
the bound organizationId.Hand off explicit follow-ups. Map specific intents to the right slash command:
/zeropath:explain <id>/zeropath:investigate <id>/zeropath:fix <id>/zeropath:triageNever invent fields. If issues.list doesn't return something
in its response shape, omit it from the table.
npx claudepluginhub zeropathai/zeropath-agent-plugin --plugin zeropathProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.