Agent skill

opencode-agent

Delegate coding tasks to an attached OpenCode server via `opencode run --attach`. Use when: (1) building new features, (2) refactoring, (3) reviewing changes, (4) iterative coding with background monitoring. This skill is server-only and always uses `--attach` with `--dir` (server-side path).

View SKILL.md on GitHub Repository
Stars 1
Forks 0

Install this agent skill to your Project

npx add-skill https://github.com/feed-mob/agent-skills/tree/main/skills/opencode-agent

Metadata

Additional technical details for this skill

openclaw 
{
    "emoji": "\ud83e\udd16",
    "requires": {
        "anyBins": [
            "opencode"
        ]
    }
}

SKILL.md

OpenCode Agent

Use this skill to run coding tasks through a remote OpenCode server.

Required Command Pattern

Always use this shape:

bash
opencode run --attach <server_url> "<task>" --dir "<server_path>"

Example:

bash
opencode run --attach http://127.0.0.1:4096 "what is in this folder" --dir "/home/richard/FeedMob"

Configuration

Read these values from environment variables:

bash
# Example values
export OPENCODE_SERVER_URL=http://127.0.0.1:4096
export OPENCODE_SERVER_DIR=/home/richard/FeedMob

# Check values
printenv OPENCODE_SERVER_URL && printenv OPENCODE_SERVER_DIR

If either value is missing, ask the user to provide the missing value(s) before running commands. Never assume, infer, or hardcode OPENCODE_SERVER_URL or OPENCODE_SERVER_DIR.

Preflight requirement before any attached run:

  1. Run printenv OPENCODE_SERVER_URL && printenv OPENCODE_SERVER_DIR.
  2. If the command fails or a value is missing/empty, ask the user for the missing value(s).
  3. Only proceed after both values are explicitly provided.
  4. Check existing projects by running opencode run --attach $OPENCODE_SERVER_URL 'pwd && ls -1d */ || true' --dir "$OPENCODE_SERVER_DIR", then list results for the user.
  5. Create a new folder for new tasks by running opencode run --attach $OPENCODE_SERVER_URL 'mkdir -p <task-folder>' --dir "$OPENCODE_SERVER_DIR".

PTY Required

OpenCode is interactive. Always run with pty:true.

bash
# Correct
bash pty:true command:"opencode run --attach $OPENCODE_SERVER_URL 'Explain async/await in JavaScript' --dir '$OPENCODE_SERVER_DIR'"

# Wrong (no PTY)
bash command:"opencode run --attach $OPENCODE_SERVER_URL 'Explain async/await in JavaScript' --dir '$OPENCODE_SERVER_DIR'"

Server-Side Directory Rule

--dir is required by this skill and points to a path on the machine where opencode server is running.

  • Do not treat --dir as a client/local path unless server and client are the same machine.
  • Prefer absolute paths for --dir.
  • The workdir field in bash does not replace --dir for attached OpenCode execution.

One-Shot Usage

bash
bash pty:true command:"opencode run --attach $OPENCODE_SERVER_URL 'Add error handling to API calls' --dir '$OPENCODE_SERVER_DIR'"

Background Usage

bash
# Start
bash pty:true background:true command:"opencode run --attach $OPENCODE_SERVER_URL 'Refactor auth module and run tests' --dir '$OPENCODE_SERVER_DIR'"

# Monitor
process action:log sessionId:XXX
process action:poll sessionId:XXX

# Respond if needed
process action:submit sessionId:XXX data:"yes"

# Stop
process action:kill sessionId:XXX

Bash Tool Parameters

Parameter Type Description
command string The shell command to run
pty boolean Use for coding agents! Allocates a pseudo-terminal for interactive CLIs
workdir string Working directory (agent sees only this folder's context)
background boolean Run in background, returns sessionId for monitoring
timeout number Timeout in seconds (kills process on expiry)
elevated boolean Run on host instead of sandbox (if allowed)

Process Tool Actions (for background sessions)

Action Description
list List all running/recent sessions
poll Check if session is still running
log Get session output (with optional offset/limit)
write Send raw data to stdin
submit Send data + newline (like typing and pressing Enter)
send-keys Send key tokens or hex bytes
paste Paste text (with optional bracketed mode)
kill Terminate the session

Rules

  1. Always use opencode run --attach.
  2. Always include --dir.
  3. Always use pty:true.
  4. Keep users updated when running background sessions (start, milestone, error, finish).
  5. If a session hangs, inspect logs first, then restart or ask user for direction.
  6. Never assume environment variable values; ask the user when missing.

Troubleshooting

bash
# Start server
opencode server --port 4096
  • Connection issue: verify OPENCODE_SERVER_URL points to a live server.
  • Path issue: verify OPENCODE_SERVER_DIR exists on the server machine.
  • No output: confirm pty:true is set and check process action:log.

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

feed-mob/agent-skills

skill-name

A clear, complete description of what this skill does and when Claude should use it

1 0
Explore
feed-mob/agent-skills

gemini-image-generator

Generate, edit, or transform images with Gemini Nano Banana using bundled Python scripts (Flash or Pro) including aspect ratio, resolution, image-to-image edits, logo overlays, and reference images. Use when users request image generation, image edits, image-to-image transformations, logo placement, or specific aspect ratios or resolutions.

1 0
Explore
feed-mob/agent-skills

feedmob-reporting-skills

MUST use this skill when using mcp__feedmob-reporting__* tools. Provides structured workflows for FeedMob reporting analysis. Key: All clients (Possible Finance, Koho Financial, TextNow, AppsFlyer MMP, etc.) use client_paid_action_count × gross_cpi (dynamic event field based on client_paid_action in click_url_histories). Critical for ensuring correct multi-step workflows and accurate data reconciliation. Trigger words: any feedmob-reporting MCP tool usage, Possible Finance, Koho Financial, TextNow, AppsFlyer, Singular reports, Adjust reports, direct spend, gross spend verification, spend reconciliation, client_paid_action.

1 0
Explore
feed-mob/agent-skills

feedmob-campaign-creator

Create FeedMob campaigns with guided workflow. Use when users request campaign creation, link generation, or new partner launches. Handles: (1) Campaign naming via feedmob-campaign-naming, (2) Client data retrieval, (3) App selection, (4) Campaign preview, (5) Campaign creation. Trigger keywords: create campaign, new campaign, generate campaign, campaign setup, link generation, partner launch, media plan campaign, 创建campaign, 生成活动, 新建campaign.

1 0
Explore
feed-mob/agent-skills

install-civitai-videoflow-bundle

Automatically installs and validates the Civitai Videoflow skill bundle from the civitai-agent-skills repository. Supports git clone or manual zip upload, runs npx skills add in dependency-safe order, and guides environment and tool prerequisite checks for CIVITAI_RECORDS_DATABASE_URL, DUOMI_API_TOKEN, IMAGEKIT_PRIVATE_KEY, and CIVITAI_ACCOUNT. Use when: user needs videoflow setup, install Civitai pipeline skills, configure image-to-video workflow, enable Civitai publish pipeline, or when civitai-videoflow and related worker skills are mentioned but unavailable. Triggers: install videoflow skills, setup civitai skills bundle, configure civitai-agent-skills, enable videoflow commands, install duomi/civitai pipeline skills.

1 0
Explore
feed-mob/agent-skills

install-weekly-hubspot-report-bundle

Automatically installs and configures weekly-hubspot-report and weekly-hubspot-report-pipeline skills from feedmob-skills repository. Supports git clone or manual zip upload, runs npx skills add commands, and guides environment variable setup for FEMINI_API_TOKEN, FEEDAI_API_TOKEN, and AWS credentials. Use when: user needs HubSpot reporting, install HubSpot skills, generate weekly ticket reports, configure report pipeline, or when weekly-hubspot-report* skills are mentioned but not available. Triggers: install HubSpot skills, setup HubSpot reporting, HubSpot weekly report, configure HubSpot, feedmob-skills installation.

1 0
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results