菜单
Use when a knowledge source is created to redact sensitive information (API keys, tokens, credentials, internal IPs, PII) and re-embed the source. Triggered automatically on knowledge.created.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Sync Fathom meetings (recorded calls + transcripts) into Hissuno as sessions. Cursor advances by `created_at`.
Register a GitHub repository as a Hissuno codebase. Parameterized: pass `repoFullName` and optional `branch`. Verifies the repo (and branch, if provided) is reachable, then POSTs to /api/codebases. Re-running is idempotent — duplicate codebases on the same (project, repo, branch) are filtered server-side.
Sync GitHub issues from a repository as feedback sessions. Parameterized: pass `repoFullName` (owner/repo) as run input. Cursor advances by issue `updated_at`. Pulls comments inline so each session has the full thread.
Handle GitHub App installation lifecycle webhooks (installation created, deleted, repositories added/removed). Records the change so the integrations UI can react. The github plugin's other webhooks (issues, pulls) should trigger their own skills as the migration progresses.
Sync Gong calls (recorded sales calls + transcripts) into Hissuno as sessions. Cursor advances by call started date.
Use when picking the next development issue to work on and implementing it end-to-end. This skill is the body of a continuous development loop - it selects the highest-priority ready issue from Hissuno, implements it, and creates a PR. Triggers on: next issue, continuous development, dev loop, what to build next, pick next task, implement next issue, development queue. Requires the hissuno CLI to be configured.
| name | hissuno-knowledge-sanitizer |
| description | Use when a knowledge source is created to redact sensitive information (API keys, tokens, credentials, internal IPs, PII) and re-embed the source. Triggered automatically on knowledge.created. |
| version | 1 |
| triggers | {"events":["knowledge.created"]} |
| input | {"sourceId":{"type":"string","required":true,"description":"ID of the knowledge source to sanitize."}} |
| capabilities | {"sandbox":true,"webSearch":false} |
| output | {"type":"object","required":["sourceId","redactions","changed"],"properties":{"sourceId":{"type":"string","description":"Knowledge source id this run sanitized."},"redactions":{"type":"integer","description":"Number of sensitive items redacted."},"changed":{"type":"boolean","description":"Whether any content was changed."},"chunksEmbedded":{"type":"integer","description":"Number of chunks re-embedded (only when changed)."}}} |
Redact sensitive information from a freshly created knowledge source and re-embed.
The harness gave you the source id (in Trigger.entity.id and as sourceId in your input) and the scope id (in Trigger.entity.snapshot.productScopeId).
The sanitization is implemented server-side. Your job is to invoke it and report the result.
hissuno sanitize knowledge "$SOURCE_ID" --scope "$SCOPE_ID" --json > result.json
The endpoint:
[REDACTED_AWS_KEY]{ sourceId, redactions, changed, chunksEmbedded? }If analyzed_content is empty (source still being analyzed or fetch failed), the response includes skipped: "no_content" and you should report redactions: 0, changed: false.
End your run with a single assistant message that includes every field in the output schema declared at the top of this skill: sourceId, redactions, changed, and chunksEmbedded if present. Do not write output.json — the harness coerces your final response into the typed object automatically.
Call report_progress with a single label like sanitize. Keep messages short.