Sentry / skills/prompt-optimizer/SKILL.md
Prompt Optimizer
Improves prompts against a small eval set: define success first, try a few candidate rewrites, keep the one that wins on held-out cases.
Source reviewed · reviewed · Named sources were inspected. No runtime results are established by this label.
01 / Task fit
Where it may help.
These are intended uses indicated by the upstream material. They are not measured outcomes.
- Prompts used repeatedly where quality can be defined
- Moving a prompt between model families
What it is
Prompt Optimizer has the agent record what a prompt is for, what good output looks like and how it fails before editing anything. If there are no examples, it builds a small eval set first. It then tries two to four candidate rewrites, compares them on the same cases, logs the changes and checks the winner on held-out cases. It covers porting prompts between model families.
Why you might use it
Use it for prompts you rely on repeatedly, such as a support reply template or a summarizer, where you can say what a good answer is. It's more work than a quick rewrite, and it's meant to be.
Put it in context.
Illustrative requests and outputs to review. These examples are separate from the recorded evaluation below.
Tighten a support reply prompt
Example request
Improve this support reply system prompt. Build 10 test cases from these anonymized tickets, compare two candidates and report which wins on 3 held-out cases.
What to look for. An optimized prompt, the eval set, an optimization log and remaining risks.
Before you rely on it. Small eval sets overfit easily. Re-check with fresh tickets before rolling it out.
Port a prompt to another model
Example request
Adapt this summarizer prompt from one model family to another and note what changed.
What to look for. A base prompt plus model-specific adapter notes.
Before you rely on it. Model snapshots change. Rerun the eval set when the model updates.
02 / Setup & access
What it asks of your environment.
Requirements
- An agent that can read the bundled reference files.
- Example inputs and success criteria, or time to build a small eval set.
Permissions & data access
- Reads the prompt and any context files you name.
- Writes prompt drafts and an eval set. It does not deploy anything.
Instructions alone do not enforce a sandbox. The agent host, connected tools and approvals determine what the skill can actually read, change or send.
03 / Things to check
Security concerns & practical limits.
These concerns follow from the reviewed sources. They describe possible exposure or required review; they are not confirmed vulnerabilities.
Review before use
Eval data can contain personal details
Test cases built from real tickets or emails can carry personal data into prompts and logs. Anonymize them first.
Limits and nuances
- Source review of SKILL.md and the repository license only. The reference files and SPEC.md were not reviewed.
- No prompt was optimized. No improvement result is claimed.
04 / Review record
What we checked. What’s still open.
Recorded review scope
- skills/prompt-optimizer/SKILL.md
- LICENSE
Not tested or established
- Source review of SKILL.md and the repository license only. The reference files and SPEC.md were not reviewed.
- No prompt was optimized. No improvement result is claimed.
- Review date
- Source revision
f57ada2d0b640406ea7e25b11d636c530e77c58b- Our relationship
- Explored. Investigated through the recorded review. No use or replication is claimed.
Upstream content can change after review. A repository revision identifies source scope; it does not prove that every script, dependency or runtime behavior was tested. How freshness is handled.
Recorded evaluation / Source review
Results with their limits attached.
Completed 4 bounded source checks on the pinned entrypoint and license. No skill workflow was executed.
- Passed
- 4
- Failed
- 0
- Total recorded
- 4
These counts refer to the cases in this report. They are not a benchmark against other skills. A source review does not establish runtime behavior.
Read the evidence reportThe report provides the case definitions, observations and baseline scope. Missing baselines or untested behaviors remain limitations.
Recorded environment
- Pinned public UTF-8 sources fetched from raw.githubusercontent.com at the recorded commit and hashed with SHA-256.
- Linux; Python 3.13.5 source-condition check. No account, credentials, private documents or live business data.
Recorded commands (2)
Commands document this evaluation. They are not setup instructions for your environment.
Read each pinned file listed in sourceManifest.python3 r6_skills.py: check each recorded source condition and reconcile report counts.Findings from this evaluation
Needs attention
Eval data can contain personal details
Test cases built from real tickets or emails can carry personal data into prompts and logs. Anonymize them first.
Observation
Stops on non-prompt problems
The skill says to stop and report when the bottleneck is model choice, retrieval or tool design rather than wording.
Evaluation limitations
- Source review of SKILL.md and the repository license only. The reference files and SPEC.md were not reviewed.
- No prompt was optimized. No improvement result is claimed.
05 / Reuse & discovery
License, attribution & popularity.
Apache-2.0
Apache-2.0 license text inspected at this revision. Keep the license and any NOTICE when redistributing covered files. The services and accounts the skill uses have their own terms.
Why it is in this collection
GitHub displayed 1,066 stars for the getsentry/skills repository on the observation date. Repository-wide; not individual skill usage, quality or safety.
1,066 repository stars
- Scope
- Entire getsentry/skills repository, not this individual skill
- Observed
- Time window
- Snapshot observed 2026-10-19
Popularity and inclusion do not establish quality, compatibility or security. Counts belong to their named metric and date; they are not a current ranking.
06 / Follow the evidence
Go to the original.
- Pinned skill instructions github.com
- Pinned license github.com
- Repository (dated star count) github.com
IonBin’s source notes are independent assessments. Upstream authors retain their work and attribution. This page does not install or execute the skill.