analyze-performance — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited analyze-performance (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Most LinkedIn AI tools that claim "analytics" either show the user numbers they could see in their dashboard, or generate generic advice based on the numbers without actually reading them. This skill exists to do the analysis work: read the data carefully, group posts by what they had in common, and identify which patterns the user should do more or less of.
LinkedIn now exposes more performance data than it used to. The Creator dashboard analytics export includes impressions, unique viewers, clicks, reactions broken out by type, comments, shares, and engagement rates per post. Plus the larger LinkedIn data archive has every post's text content.
But almost nobody actually looks at this data. They glance at impression counts, get discouraged or encouraged, and keep posting. The result is that they never learn what is working. This skill exists to do the analysis the user does not do.
Trigger when:
Do NOT trigger when:
The LinkedIn analytics export. They can get this from:
If the user has only post text and no analytics, the analysis is limited; tell them so and suggest they get the analytics export from the Creator dashboard (instant download, no waiting).
Useful additional inputs:
Do not throw averages and call it analysis. Three steps:
Group posts by:
These clusters are where the insights live. An aggregate "your average engagement rate" tells the user almost nothing. "Your carousels with 8-12 slides get 3.5x the engagement of your text posts, but your text posts with confessional hooks beat the average of both" tells them what to do.
For each cluster, look at:
What you are looking for: clusters that significantly outperform or underperform the user's overall median, in ways that are repeatable.
A pattern is only a pattern if:
Two types matter:
Be honest about both. The job is to give the user signal, not to make every post look like a learning.
# Performance analysis for [user's name or "user"]
Period analyzed: [start - end]
Posts analyzed: [N]
Median engagement rate: [N]%
Overall trajectory: [growing / flat / declining]
## What is working
### Pattern 1: [name of the pattern, e.g., "Carousels on engineering management"]
Posts in this pattern: [N]
Median engagement: [N]% (vs overall median of [N]%)
Why this is working: [one paragraph on what is specific to these posts]
### Pattern 2: ...
[2-4 strongest patterns]
## What is not working
### Pattern A: [e.g., "Posts with motivational hooks"]
Posts in this pattern: [N]
Median engagement: [N]% (vs overall median of [N]%)
Why this is likely failing: [one paragraph]
### Pattern B: ...
[1-3 weakest patterns]
## Notable outliers
### Best-performing post: [date, hook]
Why this likely worked: [analysis]
### Worst-performing post: [date, hook]
Why this likely failed: [analysis]
## Recommendations
1. [Specific, actionable. "Do more carousels on engineering management with confession-style hooks." Not "Improve your engagement."]
2. [Next]
3. [Next]
[3-5 recommendations]
## What I cannot conclude from the data
[Honest note on the limits of the analysis: small sample sizes, missing dwell time data, posts where the variable changes too much to isolate, etc.]The most dangerous outputs of content analytics are wrong patterns the user then optimizes for. Avoid:
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.