Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Show commits since the last tag in a formatted table. Use when user asks "what changed since last release", "commits since last tag", "last-tag", "what's new", or wants to see recent unreleased changes.
.claude/skills/umputun-last-tag/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | 260% | 0% |
| case-07 | ✓→✗ | ▼ Worse | -7% | 0% |
| case-15 | ✓→✗ | ▼ Worse | 46% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -21% | 0% |
| case-20 | ✓→✗ | ▼ Worse | 117% | 0% |
Show commits since the last tag in a formatted table with optional details.
Important: Avoid $() command substitution in Bash tool - use sequential steps.
bashgit fetch origin --tags
Then get the last tag:
bashgit describe --tags --abbrev=0
Store this value (e.g., v1.2.3) for use in subsequent commands.
date|author|hash|subject (substitute TAG with actual value):bashgit log TAG..HEAD --format="%ad|%an|%h|%s" --date=short
If only one unique author name appears in all rows, it's a single author.
If single author (count = 1):
Last tag: v1.2.3
Author: John Doe
| Date | Commit | Description |
|------------|---------|--------------------------------|
| 2025-12-20 | abc1234 | fix: resolve null pointer |
| 2025-12-19 | def5678 | feat: add user authentication |If multiple authors (count > 1):
Last tag: v1.2.3
| Date | Author | Commit | Description |
|------------|----------|---------|--------------------------------|
| 2025-12-20 | John Doe | abc1234 | fix: resolve null pointer |
| 2025-12-19 | Jane Doe | def5678 | feat: add user authentication |If no tag exists:
No tags found in repositoryIf no commits since tag:
Last tag: v1.2.3
No commits since this tagAfter displaying the table, use AskUserQuestion:
question: "Show commit details?"
header: "Details"
options:
- label: "All commits"
description: "Show full details for each commit"
- label: "None"
description: "Skip details"
- label: "Specific commit"
description: "Enter commit hash to inspect"If "All commits": For each commit, run:
bashgit show --stat --format="Commit: %h%nAuthor: %an <%ae>%nDate: %ad%n%n%s%n%n%b" --date=short HASH
This shows: commit hash, author with email, date, subject, body, and file change stats.
If "None": End.
If "Specific commit" or user enters hash via "Other": Run the same git show command for that commit only.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | fail→fail | 8,643 | 5,177 | -40% | 1 | 1 | 0% | 1,405 | 951 | -32% | 0 | 0 | — |
case-01 | fail→fail | 8,714 | 4,163 | -52% | 1 | 1 | 0% | 1,419 | 931 | -34% | 0 | 0 | — |
case-02 | fail→fail | 8,651 | 4,127 | -52% | 1 | 1 | 0% | 1,391 | 953 | -31% | 0 | 0 | — |
case-03 | fail→fail | 7,576 | 5,171 | -32% | 1 | 1 | 0% | 1,239 | 970 | -22% | 0 | 0 | — |
case-04 | fail→fail | 3,952 | 5,718 | +45% | 1 | 1 | 0% | 588 | 1,038 | +77% | 0 | 0 | — |
case-05 | fail→fail | 4,955 | 5,998 | +21% | 1 | 1 | 0% | 807 | 899 | +11% | 0 | 0 | — |
case-06 | fail→fail | 8,448 | 4,395 | -48% | 1 | 1 | 0% | 1,256 | 897 | -29% | 0 | 0 | — |
case-07 | pass→fail | 5,122 | 8,096 | +58% | 1 | 1 | 0% | 968 | 905 | -7% | 0 | 0 | — |
case-09 | fail→fail | 5,339 | 7,940 | +49% | 1 | 1 | 0% | 898 | 1,030 | +15% | 0 | 0 | — |
case-10 | fail→fail | 12,847 | 6,952 | -46% | 1 | 1 | 0% | 1,170 | 1,163 | -1% | 0 | 0 | — |
case-11 | fail→fail | 3,154 | 5,278 | +67% | 1 | 1 | 0% | 253 | 898 | +255% | 0 | 0 | — |
case-12 | fail→fail | 3,171 | 5,426 | +71% | 1 | 1 | 0% | 494 | 995 | +101% | 0 | 0 | — |
case-13 | fail→fail | 7,323 | 5,354 | -27% | 1 | 1 | 0% | 1,169 | 944 | -19% | 0 | 0 | — |
case-14 | fail→fail | 7,801 | 5,903 | -24% | 1 | 1 | 0% | 1,432 | 1,036 | -28% | 0 | 0 | — |
case-15 | pass→fail | 4,894 | 5,846 | +19% | 1 | 1 | 0% | 692 | 1,009 | +46% | 0 | 0 | — |
case-16 | pass→pass | 3,738 | 1,146 | -69% | 1 | 1 | 0% | 490 | 842 | +72% | 0 | 0 | — |
case-17 | pass→fail | 7,502 | 6,523 | -13% | 1 | 1 | 0% | 1,302 | 1,030 | -21% | 0 | 0 | — |
case-18 | fail→pass | 5,713 | 19,747 | +246% | 1 | 1 | 0% | 965 | 3,478 | +260% | 0 | 0 | — |
case-19 | fail→fail | 8,931 | 52,518 | +488% | 1 | 1 | 0% | 1,410 | 965 | -32% | 0 | 0 | — |
case-20 | pass→fail | 2,887 | 6,869 | +138% | 1 | 1 | 0% | 510 | 1,107 | +117% | 0 | 0 | — |
case-21 | pass→fail | 3,091 | 8,175 | +164% | 1 | 1 | 0% | 524 | 1,172 | +124% | 0 | 0 | — |
case-22 | fail→fail | 4,455 | 6,439 | +45% | 1 | 1 | 0% | 685 | 1,027 | +50% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 2 counted toward the lift figure. The other 20 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -18 percentage points is the difference between those two pass rates over the 2 comparable cases. 6 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.