- Home
- Skills
- Amnadtaowsoam
- Cerebraskills
- Summarization Rules Evidence First
summarization-rules-evidence-first_skill
- Python
1
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill amnadtaowsoam/cerebraskills --skill summarization-rules-evidence-first- SKILL.md6.4 KB
Overview
This skill provides evidence-first summarization rules that present data and metrics before conclusions to boost credibility and reduce token use. It gives concrete templates, formatting guidance, and examples for bug reports, incidents, performance issues, and status updates. The result is shorter, verifiable, and more actionable summaries.
How this skill works
The skill inspects narrative content and restructures it to lead with precise evidence: metrics, timestamps, file locations, and counts. It replaces vague adjectives with numbers, swaps general statements for specific items, and applies templates that favor lists, tables, and code references. Outputs are optimized for brevity, clarity, and verifiability.
When to use it
- Incident reports that require rapid triage and clear root-cause evidence
- Bug reports where exact error messages, frequency, and location matter
- Performance summaries and postmortems needing metrics and before/after comparisons
- Status updates for teams and stakeholders who need concise progress indicators
- Code reviews with specific file:line issues and actionable recommendations
Best practices
- Always start with evidence: metric, timestamp, count, or error string
- Prefer numbers over adjectives (e.g., '2.5s' not 'very slow')
- Be specific: include file:line, exact message, and affected scope
- Use lists or tables for multiple items and code blocks for exact snippets
- Keep action items precise and tied to expected metric improvements
Example use cases
- Turn a long outage narrative into a compact incident summary with impact, root cause, fix, and prevention steps
- Convert vague performance notes into a metrics-first brief showing before/after and recommended fixes
- Generate a bug report template populated with error, frequency, affected users, and exact location
- Produce sprint status bullets that show completed work, blocked items, and percent complete
- Create code review notes listing added/modified/removed files and file:line issues
FAQ
Generally yes; replacing prose with concise metrics and lists typically cuts tokens by ~70%, though results vary by input.
What if I don’t have precise numbers?
Flag missing data in the evidence section (e.g., 'Metric unavailable: add monitoring') and prioritize collecting exact metrics before finalizing the summary.