All skills
simota avatar

/nexus

@c805268
by shingo imotasimota/agent-skills85 stars
15

Orchestrating multi-specialist task chains and scope-adaptive product delivery: classifies intent, selects and executes the minimum viable chain, aggregates results, and verifies acceptance criteria. For multi-domain tasks, build-first delivery, and product lifecycle execution.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/nexus

This session only. Nothing lands on disk.

referencepack-subcommand.md

≈1.4k tokens on demand. Your agent reads this file only when SKILL.md points to it.

/nexus pack Subcommand

Meta-command that switches the active Claude Code skill profile by rewriting ~/.claude/settings.json skills array from a ~/.claude/profiles/<name>.json template.

Companion: _common/SKILL_PACKS.md (Pack definitions, profile catalog).

When to Use

  • User starts a new workstream and wants to scope skill listing to a domain (web / mobile / security / growth / infra / research / ai-eval / package-author).
  • The full-roster default load (90 global skills, plus any workspace-local extensions) causes routing confusion or context tax.
  • User explicitly requests /nexus pack <name>.

Forms

Form Behavior
/nexus pack Display current profile + list available profiles. No write.
/nexus pack list Same as above (explicit).
/nexus pack current Show only the active profile name.
/nexus pack <name> Switch to <name>. Confirms before writing.
/nexus pack reset Restore skills: "all".
/nexus pack restore Restore from the most recent backup.

Available <name> values

Discovered dynamically from ~/.claude/profiles/*.json. Default-shipped profiles per _common/SKILL_PACKS.md: web / mobile / security / growth / infra / research / ai-eval / package-author / skill-meta / incident-response / legal-jp / personal-env / ai-cli-admin / gemini-image / all.

Phase Contract

Phase Action
PARSE Extract subcommand form (list / current / <name> / reset / restore)
LOOKUP If <name> given, read ~/.claude/profiles/<name>.json; validate skills field is an array or "all"
BACKUP Copy ~/.claude/settings.json → ~/.claude/settings.json.bak.<unix-timestamp> (keep last 5)
DIFF Read current skills value, compute the set diff against new profile, present to user
CONFIRM AskUserQuestion — show diff summary (+N, -M skills), ask Apply? [Y/n]
WRITE On Y, replace skills field in ~/.claude/settings.json
NOTIFY Tell user: "Restart Claude Code (/exit then relaunch) for the change to take effect. Verify with /skills."

/nexus pack list and /nexus pack current skip BACKUP/CONFIRM/WRITE.

Always

  • Backup before write. Always create settings.json.bak.<timestamp> before any write. Keep the 5 most recent (delete older).
  • Validate profile JSON. Reject malformed JSON, missing skills field, or skills value not in ("all" | array of strings).
  • Preserve all other settings.json keys. Only the skills key is modified. Hooks, permissions, env, mcpServers, etc. must remain byte-identical except for that one key.
  • Confirm before write. AskUserQuestion with the +/- diff is mandatory unless user explicitly invoked with --no-confirm (CLI affordance, not implemented yet).
  • Show restart instruction. Claude Code does not hot-reload skill filters; instruct restart.
  • Use absolute path ~/.claude/settings.json, not relative.

Never

  • Modify settings.json without backup.
  • Touch any key other than skills (e.g. don't change permissions, hooks).
  • Auto-restart Claude Code (user must do it).
  • Persist a profile that fails JSON validation.
  • Skip the diff/confirm step in AUTORUN modes (this is a ## Ask First operation regardless of mode).
  • Delete ~/.claude/profiles/<name>.json files (they are durable templates).

Diff Presentation Format

## Pack switch: <current> → <new>

Adding (N):    skill_a, skill_b, skill_c, ...
Removing (M):  skill_x, skill_y, ...
Unchanged (K): [collapsed; show on request]

Apply? [Y/n]

Implementation Sketch (Internal)

The Nexus skill executes this subcommand inline (no agent spawn — single trivial settings edit per Core Rule #3 exception (a)).

Pseudocode:

profile_path = f"~/.claude/profiles/{name}.json"
profile = json.load(open(expanduser(profile_path)))
assert profile["skills"] == "all" or isinstance(profile["skills"], list)

ts = int(time.time())  # actual: from session start, not Date.now()
settings_path = expanduser("~/.claude/settings.json")
shutil.copy(settings_path, f"{settings_path}.bak.{ts}")

cleanup_old_backups(keep=5)

settings = json.load(open(settings_path))
old_skills = settings.get("skills", "all")
new_skills = profile["skills"]
diff = compute_diff(old_skills, new_skills)
# present diff via AskUserQuestion, await confirmation

settings["skills"] = new_skills
json.dump(settings, open(settings_path, "w"), indent=2)
print("Restart Claude Code to apply. Verify with /skills.")

In practice Nexus uses the Bash + Read + Edit tools to achieve the same effect (no Python required).

Edge Cases

  • Profile file missing → list available profiles and abort.
  • JSON parse error in settings.json → abort with "settings.json is malformed; fix manually before retry".
  • User declines confirmation → no write, no backup retained from this run.
  • Schema URL changed → preserve $schema field verbatim.
  • skills is currently an unknown shape (object, etc.) → treat as "all" for diff purposes; warn user.

Cross-CLI Mapping

  • Codex CLI: equivalent is ~/.codex/config.toml (no skills filter at present, ignored — this subcommand is Claude Code only).
  • agy: equivalent is ~/.gemini/antigravity-cli/settings.json (skills filter status unknown — verify before extending).

Future Work

  • --no-confirm flag for scripting.
  • Profile composition: /nexus pack web+security (union of two profiles).
  • Per-project profile (<repo>/.claude/profile.json) overriding global.
  • LEARN: track which packs the user actually uses, recommend default profile per workspace.

Source: SKILL.md on GitHub

3 warnings13d5 checks · Risk MEDIUM
  • Gen Agent Trust Hub13d

    The 'nexus' skill is a comprehensive multi-agent orchestration framework that manages complex task chains. While it incorporates extensive internal guardrails and verification protocols, it explicitly mandates the use of high-risk flags that bypass security permissions to achieve autonomy. It also provides instructions for establishing persistent tasks via cron and GitHub Actions, and utilizes external research tools to fetch content from the web.

  • Socket13d

    1 alert: gptSecurity

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    6/22 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/nexus