From 71cb051d3aa362bb61ad16fbf31c15abf554ab58 Mon Sep 17 00:00:00 2001 From: rohan-tessl Date: Mon, 30 Mar 2026 15:35:24 +0530 Subject: [PATCH] =?UTF-8?q?feat:=20improve=20agentic-mcp=20skill=20score?= =?UTF-8?q?=20(76%=20=E2=86=92=2097%)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Hey @cablate 👋 I ran your skills through `tessl skill review` at work and found some targeted improvements. Here's the full before/after: | Skill | Before | After | Change | |-------|--------|-------|--------| | agentic-mcp | 76% | 97% | +21% |
Changes made - **Description**: Replaced jargon ("three-layer progressive disclosure API") with concrete actions (start daemon, list servers, discover tools, read schemas, execute calls). Added natural trigger terms users would actually say ("connect to MCP", "run MCP tool", "check server status"). - **Eliminated redundancy**: Merged Quick Start, Core Workflow, Query Commands, and Examples sections — they all repeated the same commands. Consolidated into a single numbered workflow. - **Added validation checkpoint**: Step 2 now explicitly verifies daemon health before proceeding to queries. - **Added error recovery table**: Maps each error message to its cause and fix command, so agents can self-recover. - **Preserved progressive disclosure concept**: Token usage estimates are now inline with each workflow layer for quick reference. - **Reduced from 125 to 67 lines** while retaining all unique content.
Honest disclosure — I work at @tesslio where we build tooling around skills like these. Not a pitch - just saw room for improvement and wanted to contribute. Want to self-improve your skills? Just point your agent (Claude Code, Codex, etc.) at [this Tessl guide](https://docs.tessl.io/evaluate/optimize-a-skill-using-best-practices) and ask it to optimize your skill. Ping me - [@rohan-tessl](https://github.com/rohan-tessl) - if you hit any snags. Thanks in advance 🙏 --- SKILL.md | 110 +++++++++++++------------------------------------------ 1 file changed, 26 insertions(+), 84 deletions(-) diff --git a/SKILL.md b/SKILL.md index 3e84f3a..c6b43e3 100644 --- a/SKILL.md +++ b/SKILL.md @@ -1,65 +1,35 @@ --- name: agentic-mcp -description: Agentic MCP - Three-layer progressive disclosure for MCP servers with Socket daemon. Use when the user needs to interact with MCP servers, query available tools, call MCP tools, or manage the MCP daemon process. Provides socket-based communication for efficient server interaction with three-layer progressive disclosure API. +description: "Manage MCP servers via a Socket daemon — start the daemon, list available servers, discover tools, read tool schemas, and execute tool calls with progressive disclosure to minimize token usage. Use when the user needs to connect to MCP servers, start or stop the MCP daemon, list available MCP tools, run an MCP tool, check server status, or reload MCP configuration." --- # Agentic MCP -## Quick start +## Workflow -```bash -agentic-mcp daemon start --config # Start daemon -agentic-mcp metadata # Get server info -agentic-mcp list # List tools -agentic-mcp schema # Get tool schema -agentic-mcp call --params '{"arg":"value"}' # Call tool -agentic-mcp daemon stop # Stop daemon -``` - -## Core workflow +1. **Start daemon**: `agentic-mcp daemon start --config ` +2. **Verify running**: `agentic-mcp daemon health` — confirm `✓ Daemon is running` before proceeding +3. **Layer 1 — Server info**: `agentic-mcp metadata ` — returns name, version, status (~50-100 tokens) +4. **Layer 2 — List tools**: `agentic-mcp list ` — returns tool names + descriptions (~200-400 tokens) +5. **Layer 3 — Tool schema**: `agentic-mcp schema ` — returns full input format (~300-500 tokens/tool) +6. **Call tool**: `agentic-mcp call --params '{"argName":"value"}'` -1. `agentic-mcp daemon start` - Start daemon -2. `agentic-mcp metadata ` - Layer 1: server info -3. `agentic-mcp list ` - Layer 2: available tools -4. `agentic-mcp schema ` - Layer 3: tool details -5. `agentic-mcp call --params '{"arg":"value"}'` - Execute tool +All parameters must be a JSON object via `--params`. Append `--json` to any command for machine-readable output. ## Daemon management ```bash agentic-mcp daemon start # Start daemon +agentic-mcp daemon start --config # Start with custom config agentic-mcp daemon health # Check status -agentic-mcp daemon reload # Reload config +agentic-mcp daemon reload # Reload config without restart agentic-mcp daemon stop # Stop daemon -agentic-mcp daemon start --config # Custom config -``` - -**Sessions** (isolated instances): -```bash -MCP_DAEMON_SESSION= agentic-mcp daemon start ``` -## Query commands - -```bash -agentic-mcp metadata # Server info -agentic-mcp list # List tools -agentic-mcp schema # Tool schema -agentic-mcp daemon health # Check daemon status -``` - -## Tool calls - +**Isolated sessions** (run multiple daemons side by side): ```bash -agentic-mcp call --params '{"argName":"value"}' -``` - -**All parameters must be JSON object via `--params`**. - -**JSON mode**: -```bash -agentic-mcp metadata --json -agentic-mcp call --params '{"arg":"value"}' --json +MCP_DAEMON_SESSION=proj1 agentic-mcp daemon start +MCP_DAEMON_SESSION=proj2 agentic-mcp daemon start ``` ## Configuration @@ -76,49 +46,21 @@ agentic-mcp call --params '{"arg":"value"}' --json } ``` -## Socket protocol +## Error recovery -**Command** (newline-delimited JSON): -```json -{"id":"1","action":"metadata","server":""} -``` +| Error | Cause | Fix | +|-------|-------|-----| +| `✗ MCP daemon is not running` | Daemon not started or crashed | Run `agentic-mcp daemon start` | +| `✗ Server '' not found` | Server missing from config | Add to `mcp-servers.json`, then `agentic-mcp daemon reload` | +| `✗ Tool '' not found` | Wrong tool name | Run `agentic-mcp list ` to see available tools | +| `✗ Required argument '' not provided` | Missing param in `--params` | Run `agentic-mcp schema ` to check required args | -**Response**: -```json -{"id":"1","success":true,"data":{...}} -``` - -**Platform**: Windows (TCP) / Unix (domain socket) - -## Examples - -**Basic usage**: -```bash -agentic-mcp daemon start -agentic-mcp metadata -agentic-mcp list -agentic-mcp schema -agentic-mcp call --params '{"arg":"value"}' -``` - -**Multiple sessions**: -```bash -MCP_DAEMON_SESSION=proj1 agentic-mcp daemon start -MCP_DAEMON_SESSION=proj2 agentic-mcp daemon start -``` +## Socket protocol -## Debugging +Commands use newline-delimited JSON over Unix domain socket (or TCP on Windows): -```bash -agentic-mcp daemon health # Check status -agentic-mcp daemon reload # Reload after config change +```json +{"id":"1","action":"metadata","server":""} ``` -## Errors - -``` -✗ MCP daemon is not running -✗ Server '' not found -✗ Tool '' not found -✗ Required argument '' not provided -``` +Response: `{"id":"1","success":true,"data":{...}}`