agentleFS
zeikarzeikar/hyperclaude/.claude/skills/cc-changelog/SKILL.md

cc-changelog

Skill5 starsChanged 48 days agoMITView on GitHub

This skill should be used when the user asks to check for new Claude Code releases, "what changed in Claude Code", "any Claude Code updates", "check the changelog", "diff the Claude Code version", or wants the Claude Code changelog reviewed for changes relevant to THIS repo (the hyperclaude plugin). Reads a locally-stored last-checked version, fetches the official changelog, reports plugin-relevant changes grouped by actionability, and offers to advance the stored version.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the file is missing/empty the helper exits 2 — then ask the user which version to treat as the baseline (do not invent one) and pass it …

    line 28
  2. Very specific:

    Read the Bash output directly (no Read, no temp file):

    line 42
  3. Very specific:

    If the fetch fails (exit 1), fall back to WebFetch on https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md asking for …

    line 46
  4. Very specific:

    If the output is (up to date — nothing newer than <STORED>) (i.e.

    line 51
  5. Very specific:

    After reporting, offer to update .last-checked-version to LATEST (default: yes).

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

cc-changelog — Claude Code changelog watch (hyperclaude relevance)

Repo-local dev tool (NOT a plugin skill). Compare a stored "last-checked" Claude Code version against the current latest, mine the official changelog for everything in between, and report only what actually touches this repo's surfaces — then optionally advance the stored version so the next run diffs from here.

The value is not "summarize the changelog" — it is mapping generic Claude Code changes onto the specific surfaces of the hyperclaude plugin and verifying impact against the real repo files, exactly as a human maintainer would.

State file

.claude/skills/cc-changelog/.last-checked-version (sibling of this file, gitignored): a single line holding the last version the maintainer reviewed, e.g. 2.1.201. It is the diff baseline; it advances only in Step 6.

Procedure

Step 1 — Baseline

The diff baseline is .claude/skills/cc-changelog/.last-checked-version (one X.Y.Z line) — call it STORED. Step 2's helper reads it automatically; you don't Read it here. If the file is missing/empty the helper exits 2 — then ask the user which version to treat as the baseline (do not invent one) and pass it as the script argument.

Step 2 — Fetch and slice the changelog

The full CHANGELOG.md is ~5k lines; do not pull it whole into context. Run the local helper — it fetches the changelog and prints to stdout only the entries strictly newer than the baseline (it version-compares, so skipped numbers like 2.1.191→2.1.193 are handled; no temp file, no cleanup):

node .claude/skills/cc-changelog/get-changelog.mjs        # reads .last-checked-version
# or with a user-supplied baseline: node .claude/skills/cc-changelog/get-changelog.mjs 2.1.201

Read the Bash output directly (no Read, no temp file):

  • line 1 LATEST=<x.y.z> — the topmost changelog version → this is LATEST.
  • the rest — the sliced entries, verbatim, newest first (or (up to date — …) if none).

If the fetch fails (exit 1), fall back to WebFetch on https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md asking for entries strictly greater than STORED.

Step 3 — Up-to-date short-circuit

If the output is (up to date — nothing newer than <STORED>) (i.e. LATEST == STORED, nothing newer), report "already current at STORED, nothing new" and STOP. Do not rewrite the state file.

Step 4 — Map each entry to a hyperclaude surface

For every entry newer than STORED, decide which surface (if any) it touches, then classify into one of four buckets. Report a change only if it lands on a surface; drop pure end-user-UI/terminal/voice items unless they affect this repo.

Buckets:

  • 🔧 Actionable — needs a change in this repo (a skill/agent/hook/doc edit).
  • ✅ Favorable — the platform now does natively what the plugin did manually; no action, but note it as a future-simplification candidate.
  • 🟢 Safe / no-op — plausibly relevant but verified harmless for this repo.
  • ℹ️ Informational — worth knowing, behavior-neutral.

Surface checklist (what to grep/read to confirm impact — never assert impact without checking the actual file):

SurfaceWhat to checkRepo anchors
Hookshooks.json matcher semantics (hyphen/comma/regex), SessionStart & PostToolUse events, hook stderr/exit-code behaviorhooks/hooks.json, hooks/*.mjs
Loop spawn/reply transportbackground-agent spawn semantics for a no-name: Agent call, whether the final text still arrives as the task-notification <result>, SendMessage routing by returned agentId, subagent prompt-cache TTL, and any change to plugin-agent definition resolution (upstream #78234 / #81746)references/loop-protocol.md, skills/hyper-*-loop/**
Agent dispatch (non-loop)run_in_background default, subagent depth cap, Agent(type) deny/allow rules, subagent model inheritanceskills/hyper-implement/SKILL.md, skills/hyper-plan, hyper-research, hyper-docs-sync
Skill/agent frontmattername/description/metadata.* parsing, kebab/snake/camel, malformed-YAML handling, slash-command loadingskills/**/SKILL.md, agents/*.md
Codex bridge / Bash spawnanything changing how Bash spawns or sandboxes child processes (Codex is external, but the bridge is spawned via Bash)scripts/codex-bridge.mjs, scripts/codex/*.mjs
Plugin manifest / validate / releaseclaude plugin validate, plugin.json, marketplace/rename behavior.claude-plugin/plugin.json, docs/development.md
Overlapping built-insbuilt-in /code-review, /agents wizard, default model, /review — usually informational, but flag stale doc referencesdocs/**, README.md

For any flagged-but-uncertain item, deepen before concluding: consult https://docs.claude.com/en/docs/claude-code (WebFetch) or WebSearch, and grep/read the anchor files to confirm the actual impact.

Step 5 — Report

Lead with a one-line TL;DR answering "does anything need action?". Then the four buckets (omit empty ones or say "none"). Cite repo files as clickable [path](path) links. Keep it tight — this mirrors a maintainer triage, not a full changelog reprint.

Step 6 — Advance the baseline

After reporting, offer to update .last-checked-version to LATEST (default: yes). On confirmation, overwrite the file with LATEST + newline. Skip the bump if the user wants to keep the current baseline for re-review. Never bump before reporting.

Notes

  • Primary source is the CHANGELOG (authoritative per-version); official docs are a secondary deepening source for a specific flagged feature.
  • This skill only reads and reports — it never edits plugin code as a side effect. Any actionable finding is handed back to the user to act on (e.g. via hyper-plan).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.