Agents Md
Generate and align a canonical AGENTS.md plus tool-call mappings so Antigravity CLI agents know which workspace tools replace familiar Read/Write/Bash/Skill patterns.
Overview
agents-md is an agent skill for the Build phase (agent-tooling) that documents canonical AGENTS.md usage and Antigravity CLI tool-call equivalents for standard coding-agent tools.
Install
npx skills add https://github.com/devarfeen/agent-skills-kit --skill agents-mdWhat is this skill?
- Maps 11 common agent tools (Read, Write, Edit, Bash, Grep, Glob, TodoWrite, Skill, WebSearch, WebFetch, Task) to Antigra
- Documents @ file references, parallel subagents via start_subagent, and optional isolated Git worktrees
- States Antigravity reads AGENTS.md from the active workspace with no Antigravity-specific shim in this kit
- Explicit memory guidance: no repo memory files or memory MCP servers—use generated AGENTS.md and binding context only
- Points to tool-calling.md and memory guidance for full kit behavior
- 11 tool mappings listed in the tool-calling index table
Adoption & trust: 1 installs on skills.sh; 3/3 security scanners passed (skills.sh audits); trending (+100% hot-view momentum).
What problem does it solve?
You moved to Antigravity CLI but your prompts and habits still assume Claude Code tool names, so agents mis-call tools or invent memory patterns the kit does not support.
Who is it for?
Solo builders adopting devarfeen’s agent-skills-kit with Antigravity CLI who want one canonical instruction file and a clear tool-name cheat sheet.
Skip if: Teams that do not use Antigravity CLI or who need persistent cross-repo memory stores this kit explicitly discourages.
When should I use this skill?
Setting up or updating Antigravity CLI workspace context with AGENTS.md from the agent-skills-kit.
What do I get? / Deliverables
After applying the kit guidance, your workspace ships a consistent AGENTS.md and tool-calling index so agents invoke the right CLI commands and avoid unsupported memory-file hacks.
- Aligned AGENTS.md convention
- Tool-name mapping reference for agents
- Memory policy guidance (no repo memory files)
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Agent instruction files are created and maintained while you wire up coding agents and their workspace bindings, which is core Build-phase agent-tooling work. The kit documents Antigravity CLI equivalents for standard agent tools and treats AGENTS.md as the primary context file—exactly what solo builders configure under agent-tooling.
Where it fits
Scaffold AGENTS.md so a new repo’s Antigravity sessions know read_file replaces Read and run_command replaces Bash.
Document WebSearch/WebFetch mappings before wiring research scripts into the same workspace.
Verify onboarding docs match the kit’s tool table before handing the repo to a contractor agent.
Update binding context when you add browser_subagent flows without introducing forbidden memory files.
How it compares
Reference meta-skill for workspace agent conventions—not an MCP server or a task automation integration.
Common Questions / FAQ
Who is agents-md for?
It is for indie developers and small teams standardizing Antigravity CLI workspaces with the agent-skills-kit, especially when migrating from Claude Code-style tool vocabulary.
When should I use agents-md?
Use it during Build (agent-tooling) when you first scaffold AGENTS.md, when you add parallel subagents, and again in Operate (iterate) when you refresh bindings after tool or workflow changes.
Is agents-md safe to install?
It is documentation-oriented within the kit; review the Security Audits panel on this Prism page and inspect generated AGENTS.md before granting shell or network permissions to your agent.
SKILL.md
READMESKILL.md - Agents Md
Tool-calling index: [`tool-calling.md`](tool-calling.md). | Skill Reference | Antigravity CLI Equivalent | | :--- | :--- | | `Read` (file reading) | `read_file` | | `Write` (file creation) | `write_file` | | `Edit` (file editing) | `replace` | | `Bash` (run commands) | `run_command` | | `Grep` (search content) | `grep_search` | | `Glob` (search by name) | `glob` | | `TodoWrite` (task tracking) | `write_todos` | | `Skill` tool (invoke a skill) | auto-activated from `SKILL.md` metadata (no explicit tool; mention skill name to force activation) | | `WebSearch` | `search_web` | | `WebFetch` | `read_url_content` | | `Task` tool (dispatch subagent) | `start_subagent` (proto field `invoke_subagent`); `browser_subagent` for browser tasks | **Key Notes:** - `@` references files/context (e.g. `@src/main.go`), not agents. - Parallelism comes from the Agent Manager and `start_subagent`: the orchestrator decomposes a goal and spawns dynamic subagents that can share the parent's workspace or run in an isolated Git worktree (clean context window, same model). - Antigravity CLI reads `AGENTS.md` directly from the active workspace as a supported context file. This kit treats `AGENTS.md` as the canonical instruction file for Antigravity CLI and emits no Antigravity-specific shim. - **Memory:** No native memory file store in this kit's supported model. Use generated `AGENTS.md` and binding context files. Do not create repo memory files or add memory MCP servers for this kit. See [`memory-global-defaults.md`](memory-global-defaults.md). - Multi-repo workspace policy: use workspace-root MCP config. - For exact config-file placement by tool, use [`tool-calling.md`](tool-calling.md). ## Agents: parallel, background & roles Parallelism is built into the **Agent Manager**: `start_subagent` spawns dynamic, dependency-aware subagents (parallel execution; specific concurrency limit is not publicly documented). Local background: `/schedule` runs cron-style tasks that survive app close, and Artifacts (plans, diffs, walkthroughs) are written to a folder you review on return. **Managed Agents API / remote managed execution** is cloud — do not use it. | Role | Antigravity mechanism | | :--- | :--- | | Orchestrator | lead agent / Agent Manager (owns merge + final judgment) | | Explorer | `start_subagent` with a read-only subagent definition | | Researcher | `start_subagent` + `search_web` / `read_url_content` | | Planner | planning mode (task groups, Artifacts) | | Implementer | `start_subagent` with a write-enabled subagent definition | | Reviewer | `start_subagent` with a read-only reviewer subagent definition | | Tester | `start_subagent` running tests / build | | Tool-runner | `start_subagent` scoped to shell; `browser_subagent` for browser sequences | MCP: project `.agents/mcp_config.json`; remote HTTP entries use `serverUrl`. Skills: project `.agents/skills/<name>/SKILL.md`; user-global `~/.gemini/antigravity/skills/<name>/` (also `.agent/skills/` accepted as a back-compat path). ## Permissions, hooks, slash commands - **Permission resources** (config + `/permissions`): `read_file`, `write_file`, plus MCP-tool filtering. Rules use `action(target)` form with Allow / Deny / Ask lists. - **Permission modes**: `request-review` (default), `proceed-in-sandbox`, `always-proceed`, `strict`. - **Highest elevated permission launch**: `agy --dangerously-skip-permissions`. Do not combine it with `--sandbox` when the goal is full elevation; `--sandbox` enables terminal restrictions. - **Hooks**: `PreToolUse` / `PostToolUse` with regex `matcher` on tool name; JSON schema includes `toolCall.name`, `toolCall.args`, `stepIdx`, plus common fields (`conversationId`, `workspacePaths`, `transcriptPath`, `artifactDirectoryPath`). Hook decision values: `allow`, `deny`, `ask`. - **Useful slash commands**: `/goal`, `/grill-me`, `/schedule`, `/browser`, `/artifact`, `/permissions`, `/context`, `/btw`, `/model`, `/config`. - **Non-interactive**: `agy -p "<prompt>"` for pi