dd_priorities ยท diff
git:20260906.f9cef19 to git:20260925.f86d8db
14 added, 6 removed. Audit A to A.
---
name: dd_priorities
- description: Synthesize a saved DD checklist report into up to eight decision-relevant priorities. Use after dd_checks when the most material concerns and follow-up actions are needed.
+ description: Merge the most important findings of a saved DD checklist report into up to five decision-relevant priorities, including risks that grow when findings occur together. Use after dd_checks when the most material concerns and follow-up actions are needed.
---
# DD priorities
- Prioritize the concerns and evidence gaps in an existing DD report.
+ Prioritize the most important concerns and evidence gaps in an existing DD report.
## Inputs and outputs
The async `dd_priorities(startup)` returns one `InsightFile` in a list,
named `dd-priorities-<startup>-<model>.md`.
## Workflow and dependencies
Resolve the canonical startup slug and read the preferred stored `dd_checks`
report with `find(selection="any")`. Input selection does not establish freshness.
The registry declares `dd-checks`; direct calls never rerun it.
+ Take the report's most important findings through
+ `dd_checks.most_important_findings`. These are the checks with the highest
+ importance, sorted by it. A report without that section, such as a manual one,
+ is used completely.
+
Read the report before checking manual-first output reuse. The cache covers
- indexed startup revisions, effective synthesis instructions and the actual report
- content. On a miss, send the complete report to `generate_markdown`.
+ indexed startup revisions, effective synthesis instructions and the selected
+ findings. On a miss, send the findings to `generate_markdown`.
- The prompt asks for up to eight distinct concerns with supporting checklist IDs,
- citations and follow-ups. These semantic requirements are not mechanically verified.
+ The prompt asks for up to five distinct priorities: merge findings about the
+ same issue, name combinations that make a risk larger, and order the result by
+ importance for the investment decision. Each priority keeps supporting check
+ IDs, citations and follow-ups. These semantic requirements are not
+ mechanically verified.
## Side effects and failure behavior
Read stored inputs, call a model when needed and save the synthesis. No discovery,
source import, semantic search or ranking engine is involved. Missing or empty
DD reports raise with instructions to run DD checks; generation failures propagate.
## Usage
```bash
conda run -n sictic-env python -m skills.harness '/dd_priorities "<STARTUP>"'
```
The direct CLI uses `--startup`.
## References
- [Implementation](dd_priorities.py)
- [Synthesis instructions](../../config/dd_priorities/llm_instructions.md)
- [Source workflow](../dd_checks/SKILL.md)