Sem inspect weave
Semantic version control workflow: sem (structural truth) -> inspect (review prioritization) -> weave (coordination + merge semantics).From its SKILL.md
npx -y skills add timseriakov/sem-inspect-weave-skill --skill sem-inspect-weaveAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
- runs commandsInstructs the agent to run 7 commands, including `sem diff --from HEAD~1 --to HEAD --format json` and 6 more.
SKILL.md
3.2 KB, 612 tokens by cl100k_base, as published. Nobody here has run it
Sem / Inspect / Weave
This skill teaches a single coherent workflow:
sem -> inspect -> weave
Use it when you need entity-level understanding (functions/classes/properties), risk-focused review prioritization, and safe merge/coordination semantics.
When to Use
- Semantic or entity-level diff analysis (not just line diffs)
- Impact questions (what else breaks if this entity changes?)
- Risk triage (what deserves careful review first?)
- Branch integration prep / semantic conflict analysis
- Multi-agent overlap (avoid false conflicts)
When Not to Use
- Trivial changes where direct reading is cheaper
- Any workflow that would mutate repo merge configuration by default
Reasoning Model
- Discover changes -> sem
- Understand impact -> sem impact / sem graph / sem blame
- Prioritize review -> inspect
- Coordination + merge semantics -> weave
Tool Roles
| Tool | Core Role | Notes |
|---|---|---|
sem | structural truth | Entity-level diff/graph/impact/blame. Establishes what changed. |
inspect | review prioritization | Classifies + risk-scores entities (blast radius, dependents, public API) and groups changes for review order. |
weave | coordination + merge semantics | Entity-level merge semantics and optional advisory coordination (claims). |
Trigger Phrases
- semantic diff
- entity-level diff
- use sem to analyze changes
- review this change semantically
- analyze impact of this change
- find risky entities in this diff
- prioritize review for this commit
- prepare safe merge
- detect semantic conflicts
Typical Workflows
Review last commit
sem diff --from HEAD~1 --to HEAD --format json
inspect diff HEAD~1 --format json
Review branch vs main
sem diff --from origin/main --to HEAD --format json
inspect diff origin/main..HEAD --format json
Analyze refactor impact
sem impact <entity>
sem graph
Prepare merge safely
weave preview main
Weave Safety Rules
Run weave only when:
- merge conflicts exist
- multiple agents/branches touched the same entities or adjacent semantic areas
- preparing branch integration
- semantic conflict context is explicitly needed
Never run weave setup unless the user explicitly asks (it mutates repo merge-driver configuration).
Wrapper Policy
Scripts in scripts/ are optional helpers.
- You may call
sem,inspect, andweavedirectly. - Use wrappers only to standardize structured output and reduce boilerplate.
Output Contract
- Structural changes (entities) from sem.
- Review hotspots and risk ranking from inspect.
- Merge/coordination risks from weave (or explicitly state weave was not needed).
What ships with it: 11 files
8.3 KB alongside SKILL.md, 4 of them executable
agents/
- openai.yaml270 B
references/
- commands.md734 B
- decision-guide.md492 B
- examples.md446 B
- inspect-risk-model.md428 B
- limitations.md402 B
- weave-coordination.md610 B
scripts/
- inspect-review-targetruns947 B
- sem-jsonruns735 B
- skill-preflightruns3.1 KB
- weave-preview-saferuns227 B