Audit outbound replies against original multi-part requests for exact match.
npx skills add https://github.com/dimagi-internal/ace --skill self-review-dimagi-internal
Please help me install this Agent Skill. Skill: self-review Source: https://github.com/dimagi-internal/ace/tree/main/skills/self-review Command: npx skills add https://github.com/dimagi-internal/ace --skill self-review-dimagi-internal
This Skill solves the frequent issue of outbound replies and counterpart-facing deliverables that fail to fully address the requester's original asks, leading to rework, miscommunication, and missed requirements. Near-miss responses (like linking a summary instead of the requested report) and dropped items from multi-part requests are common failure points this Skill eliminates.
Run the self-review skill to audit your draft reply to the stakeholder's multi-part request before sending it to confirm every ask is addressed exactly as requested.
High-intent search queries and answers about installing and using this skill.
To verify an outbound reply matches a multi-part request, extract every discrete ask verbatim into a checklist, then confirm the deliverable addresses each item exactly as requested before sending to eliminate dropped items.
Deliverables fail to address requests due to near-miss responses and dropped items, often caused by substituting summaries for exact links or overlooking multi-part asks; auditing deliverables against verbatim extracted asks eliminates these gaps.
The best way to audit a draft response is to perform direct source verification by opening cited documents, rate the deliverable on faithfulness and completeness using a 1-5 scale, and fix all identified gaps prior to sending.
To check if cited sources are accurate, open and read the linked documents or run artifacts directly instead of relying on memory, ensuring the cited information exactly matches the content within the original source materials.
Yes, you can use an automated review process for high-stakes deliverables with linked documents to enforce verbatim ask extraction, flag deliberate substitutions, and require all gaps to be fixed before the correspondence is sent.
kbanc85Ingest and consolidate source documents into a unified memory with provenance.
iPythoningSchedule and manage delayed segmented message deliveries across WhatsApp, Telegram, and email.
PolliticalSolutionsExtract diligence issues from VDR documents with materiality thresholds and house memo outputs.
Leep-GHScan open issues and discussions to classify draft responses for human review.
