with one click
changelog
Generate a structured changelog entry from git changes
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Menu
Generate a structured changelog entry from git changes
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Produce demo videos of Explorbot runs for social media and presentations — composites the real browser screencast with a simulated terminal replaying real log lines in near real time. Use when asked to create a demo video, promo clip, or presentation video of an Explorbot session.
Diagnose a failed Explorbot session via /explorbot-debug, then propose ONE minimal fix to pilot/tester. Strongly biased toward prompt-only changes; rejects over-engineering.
Audit AI prompts and rules for web navigation and testing. Use when reviewing src/ai/rules.ts, navigator.ts, or tools.ts for contradictions, gaps, locator guidance, or tool/schema consistency.
Debug failed Explorbot interactions. Analyzes Langfuse traces or log files to find why tests failed and suggests Knowledge fixes.
Interactive XPath testing against HTML files in test-data/. Find elements by iteratively querying XPath expressions.
Process images for documentation - add borders/shadows to screenshots, create GIFs from videos. Use when preparing visual assets.
| name | changelog |
| description | Generate a structured changelog entry from git changes |
| user_invocable | true |
Generate a structured changelog entry for Explorbot by analyzing git changes.
/changelog — generate from all uncommitted changes (staged + unstaged + untracked)/changelog HEAD~5 — generate from last N commits/changelog v1.0..HEAD — generate from a ref rangeSTOP. Before writing a single line of changelog, you MUST run
git diffover the full scope this skill requires. Do not infer changes from recent commits, from the conversation history, or from memory of what you just edited. The working tree is the source of truth. Skipping this step produces wrong, incomplete changelogs — every time.
This applies even if you just committed: a fresh commit is a strict subset of the working tree when unstaged changes exist. Always run the full scope below.
Run ALL of these, in parallel, every time:
git status --short — see what's modified, staged, and untrackedgit diff --stat HEAD — combined stat for every tracked file vs HEAD (staged + unstaged)git diff HEAD <file> <file> … — the actual diff for every file that appeared in step 2, batched across the key paths listed belowRead its full contents (they have no diff)/changelog <ref>)git diff <ref> --stat — stat across the ref rangegit diff <ref> <file> <file> … — the actual diffsIf any of these paths appears in the stat, you MUST read its diff before writing the entry:
bin/explorbot-cli.ts — CLI entry pointsrc/commands/ — every file (new TUI commands, new flags)src/config.ts — config surfacesrc/ai/tools.ts — tool descriptions the AI seessrc/ai/rules.ts — rules injected into promptssrc/ai/*.ts — agent behaviorsrc/explorer.ts, src/explorbot.ts — container + orchestrationsrc/reporter.ts — reporter output users seesrc/state-manager.ts — state + transitionssrc/experience-tracker.ts — experience file formatsrc/stats.ts — user-visible statsdocs/ — docs changesBefore you start writing, answer (internally) these two questions:
git status --short AND git diff --stat HEAD in this turn?If either answer is no, go back and run the commands. Do not continue.
Map changed files to changelog sections:
| Source | Section |
|---|---|
bin/explorbot-cli.ts — new .option() calls, new commands | New CLI Options |
src/commands/ — new flags parsed from args | New TUI Commands |
src/config.ts — new fields in config interfaces | Configuration |
src/ai/*.ts — behavior changes in agent classes | Agent-related changes |
Include:
Skip:
Use today's date as the heading. Follow this format exactly:
## YYYY-MM-DD
### New CLI Options
- **`--flag-name`** — Description of what it does.
```bash
explorbot command --flag-name # example usage
explorbot command --flag-name value # with value
/command --flag — Description. Document with TUI-specific syntax.
/command --flag
/command arg --flag
config.key — What it controls. Default: value.nested.parent.key — Description. Default: value.
**Rules:**
- Only include sections that have entries (omit empty sections)
- Every CLI/TUI option must have a usage example in a code block
- Every CLI/TUI option must explain what it does from the user's perspective — not just mention the flag name
- If a command gains new flags (e.g. via shared options helper), list each new flag separately with its description
- Configuration entries list the key path, description, and default value — no code blocks
- For agent-related changes, prefix with `[AgentName]` in square brackets
- For non-agent components (State Manager, Experience Tracker), use plain name prefix
- Only list changes with user-visible behavior impact
- NEVER use internal developer terminology — describe what the user sees or experiences, not implementation details
- Do not invent jargon like "gap-fill" or "annotation pipeline" — use plain language ("elements that AI missed are now detected")
## Step 5: Write to CHANGELOG.md
- If `CHANGELOG.md` doesn't exist, create it with a `# Changelog` header followed by a blank line, then the new entry
- If `CHANGELOG.md` exists, prepend the new entry after the `# Changelog` header line (newest first)
- Always leave a blank line between the header and the first entry, and between entries