agentconf

Plugin agentconf from marketplace bytedesk in ByteDeskAI/bytedesk-marketplace. Commands read from the repo's manifest — 2 days ago.

Skills (9)

• adopt

  Take a machine that already has per-tool instruction files and work out what belongs in the shared source, before wiring anything. Use on first setup, when the user already maintains CLAUDE.md and AGENTS.md by hand, or when they ask how to start using agentconf without losing what they have.

• audit

  Add an unmanaged AI coding tool to agentconf by discovering how it loads configuration and proving it with a canary. Use when a tool is installed but has no adapter, or when it is unclear whether a tool reads a given config path.

• doctor

  Report the state of shared agent configuration on this machine — which agent tools are installed and logged in, whether the shared instruction source is wired into each, and whether anything has drifted. Use when the user asks about agent config, AGENTS.md, CLAUDE.md, whether a tool is picking up th

• providers

  Report which LLM providers are configured on this machine, which tools use them, and where two auth signals conflict — such as an API key silently overriding a subscription login. Use when the user asks what providers or models are set up, why usage is billing to the wrong account, or whether a tool

• restore

  Undo an agentconf change by putting back a backup it took before writing. Use when a wire went wrong, when a tool started behaving differently after a config change, or when the user wants their previous instruction file back.

• scan

  Find every AI coding agent installed on this machine and report which are managed, which are installed but unproven, and which are unrecognised. Use when the user asks what agents are on their system, whether a newly installed tool is covered, or why a tool is not picking up their shared rules.

• setup-cli

  Put the agentconf command on the user's own PATH, so it can be run from a terminal or by another agent rather than only from inside Claude Code. Use after installing agentconf, when `agentconf` returns command not found in a shell, or when the user wants Codex or a script to invoke it.

• verify

  Prove that each agent tool actually reads the shared instruction file, using a canary token rather than assuming the write landed. Use when a tool seems to be ignoring its instructions, after adding a new adapter, or when the user asks whether their rules are really in effect.

• wire

  Set up or repair the shared agent configuration — one instruction source at ~/.agents/AGENTS.md wired into every installed agent tool. Use when the user wants to stop hand-syncing CLAUDE.md and AGENTS.md, when `agentconf check` reports something broken, or after an editor has clobbered a managed sym

Subagents (1)

• config-auditor

  Discovers how an unmanaged AI coding tool loads its configuration, and produces an adapter only if a canary proves it. Use when adding a new tool to agentconf, or when a tool's config location is uncertain.

Hooks (2)

• SessionStart

  agentconf check --quiet --context

• PostToolUse

  agentconf check --quiet --context --if-touched

Similar plugins