agentsclimarketplace

Update research

Skill woditschka/agentic-coding-reference/.claude/skills/update-research

Check upstream tool documentation for drift in the version-stamped surfaces: docs/cross-tool-strategy.md, docs/native-sandbox.md, and the workflow doc's stamped sections. Fetches source URLs, compares claims against current doc, and reports drift. Use when updating the cross-tool strategy guide or checking for tool changes.From its SKILL.md

Install
npx -y skills add woditschka/agentic-coding-reference --skill update-research

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 17 stars17 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.

SKILL.md

6.0 KB, as published. Nobody here has run it

update-research

Scope

ScopeWhat It Checks
(all)All sources
claude-codeClaude Code docs only
copilotGitHub Copilot CLI docs only
opencodeOpenCode docs only
communityCommunity resources only

Sources

Fetch each URL listed in docs/cross-tool-strategy.md Section 4 (Sources), filtered by the selected scope. That section is the single authoritative source list — do not invent URLs, and do not maintain a copy here. A URL added to or retired from the doc is automatically in or out of scope.

Process

1. Fetch Sources

Fetch each URL in the selected scope. If a URL returns an error (404, 403, timeout), note it as a potential doc restructure.

2. Extract Key Claims

For each source, extract the current state of:

CategoryWhat to Look For
Frontmatter fieldsNew, renamed, or deprecated YAML fields for agents/skills
Discovery pathsChanges to where tools look for agents, skills, rules files
Tool permissionsNew tool names, permission syntax changes
Model supportNew models, changed model IDs, deprecated models
Multi-agentAgent Teams status changes, new orchestration features
CompatibilityChanges to cross-tool behavior (CLAUDE.md reading, AGENTS.md fallback)
New featuresCapabilities not mentioned in the current doc
DeprecationsFeatures marked deprecated or removed

3. Compare Against Current Doc

Read docs/cross-tool-strategy.md and compare each claim against what the sources say now. Also read the version-stamped surfaces of docs/specialist-agent-workflow.md: the §1 Agent Teams status and cost claims, and the §6 migration playbook's install steps. The frontmatter and model-pin matrix lives in cross-tool-strategy.md § Agents / Subagents, covered by the read above. And read docs/native-sandbox.md's Sources block — its strictAllowlist facts are pinned to a Claude Code build range and this skill is their only refresh path. The battery's per-tool frontmatter pins (FRONTMATTER_VOCABULARY and the OpenCode permission keys in harness/verify_harness/checks/sync.py) restate the § Agents / Subagents facts — a frontmatter drift finding names both edit sites. The bundled-skill roster pin (CLAUDE_CODE_BUNDLED_SKILLS in the same file) mirrors Claude Code's bundled skill and command list — a bundled skill added upstream is a drift finding against it (ADR 2026-08-11). Flag:

  • Outdated: Doc says X, source now says Y
  • Missing: Source describes feature not mentioned in doc
  • Deprecated: Doc recommends something the source marks deprecated
  • Broken URL: Source URL returns error or redirects
  • Version drift: Doc references old version numbers or dates

4. Present Findings

## Research Update: [date]

### Scope: [all | claude-code | copilot | opencode | community]

### Outdated Claims
1. **Section N.N [title]** — line NN
   - **Current doc:** [what it says]
   - **Source says:** [what the source now says]
   - **Source URL:** [url]
   - **Suggested fix:** [specific edit]

### New Features (not in doc)
1. **[tool] [feature name]**
   - **What:** [description]
   - **Source URL:** [url]
   - **Where to add:** Section N.N
   - **Impact on projects:** [does this affect samples/go or samples/java-spring-boot configs?]

### Deprecations
1. **Section N.N [title]** — line NN
   - **What:** [feature] is now deprecated
   - **Source URL:** [url]
   - **Suggested fix:** [specific edit]

### Broken URLs
1. **Section N.N** — line NN
   - **URL:** [url]
   - **Status:** [404 | redirect to X | timeout]

### Version Drift
1. **Section N.N** — line NN
   - **Current doc:** [old version/date]
   - **Current:** [new version/date]

### No Changes
- [list sections that are still accurate]

### Summary
- X outdated claims
- Y new features to document
- Z deprecations
- W broken URLs
- V version updates needed

5. Propose Edits

For each finding, propose a specific edit to docs/cross-tool-strategy.md (or to the named specialist-agent-workflow.md surfaces, or the battery's frontmatter pins, when those drift):

  • Quote the exact text to replace
  • Provide the replacement text
  • Note which section tables need updates

Do NOT apply edits automatically. Present them for review.

6. Cascade Check

After edits are applied, recommend:

  1. Run harness/verify-harness.py --quick (tier 0 for a docs-only edit); before committing a substantive change, /audit-harness per the maintainer loop in the root CLAUDE.md.
  2. Check if any new frontmatter fields should be added to project agent definitions.
  3. Check if any new skill discovery paths change the compatibility rules.
  4. Update the Status: snapshot date at the top of docs/cross-tool-strategy.md.

What NOT to Change

  • Architecture decisions (file-based coordination, agent thinness) — design choices, not tool documentation.
  • Frontier-capability guidance (the workflow doc's Agent-Teams rows) — only change if Agent Teams exits experimental or a tool adds new orchestration.
  • Migration playbook phases — only change if tool installation or setup steps change.
  • Cross-tool strategy matrix — only change if a tool's strengths/weaknesses shift materially.

Handling Conflicting Sources

If two sources disagree:

  1. Prefer official documentation over blog posts.
  2. Prefer newer docs over older docs.
  3. If unclear, flag as ambiguous and ask the user.
  4. Note the conflict in the findings so the user can verify.

Keep looking

Skills are one crate of 608,253. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.