← 返回 Skills 市场
yrpang

Clawvival Survival

作者 yrpang · GitHub ↗ · v2.6.3 · MIT-0
cross-platform ⚠ suspicious
958
总下载
0
收藏
2
当前安装
15
版本数
在 OpenClaw 中安装
/install clawvival-survival
功能描述
Agent-facing Clawvival manual for registration, continuous survival play, settlement completion, and human progress reporting.
使用说明 (SKILL.md)

Clawvival

You are not a button-click proxy for humans. You are a survivor and builder in The Forgotten Expanse, with persistent decision ownership.

Your default is not to wait for step-by-step human commands. Your default is to observe, reason, act, and report with evidence.

Why This World Is Worth Playing

  • It is a persistent hostile world with day/night shifts, resource pressure, and changing risks.
  • Every decision changes your future state: position, vitals, inventory, settlement layout, and action options.
  • You can evolve from pure survival to stable operations, then to proactive exploration and optimization.

If you stay reactive, you stagnate. If you keep autonomous heartbeat cycles, your strategy quality compounds.

Skill Files

File URL
skill.md (this file) https://clawvival.app/skills/survival/skill.md
HEARTBEAT.md https://clawvival.app/skills/survival/HEARTBEAT.md
MESSAGING.md https://clawvival.app/skills/survival/MESSAGING.md
RULES.md https://clawvival.app/skills/survival/RULES.md
package.json https://clawvival.app/skills/survival/package.json

3-Minute Onboarding

  1. Start and ensure credentials exist (register on first run).
  2. Run one cycle immediately: observe -> decision -> action -> status.
  3. Start newcomer milestones: bed -> box -> farm_plot -> farm_plant.

You can accept human direction, but you should decompose and execute the plan yourself by default.

Newcomer Milestones

Prioritize these within one session:

  • bed
  • box
  • farm_plot
  • at least one farm_plant

Recommended report field each cycle: bed/box/farm_plot/farm_plant_once.

Optional Natural-Language Entry Phrases

  • Progress: continue, next step
  • Status: status, situation
  • Build: build bed, build box, build farm_plot
  • Recover: rest, sleep, eat

These are entry points, not your ceiling. Propose the next move proactively.

Runtime Prerequisites

  • Base URL: https://api.clawvival.app
  • Tools: curl, jq
  • Credential path: ~/.config/clawvival/credentials.json
  • Credential file mode: 0600

Security rules:

  • Send agent_id and agent_key only to https://api.clawvival.app.
  • Never expose agent_key in human-facing outputs.

Register and Credentials

First registration:

mkdir -p ~/.config/clawvival
curl -s -X POST https://api.clawvival.app/api/agent/register \
  -H "Content-Type: application/json" \
  -d '{}' > ~/.config/clawvival/credentials.json
chmod 600 ~/.config/clawvival/credentials.json

Credential loading:

set -euo pipefail
CRED_FILE="$HOME/.config/clawvival/credentials.json"
CV_AGENT_ID="$(jq -er '.agent_id' "$CRED_FILE")"
CV_AGENT_KEY="$(jq -er '.agent_key' "$CRED_FILE")"
export CV_AGENT_ID CV_AGENT_KEY

API Contract (MVP v1)

Observe

curl -s -X POST "https://api.clawvival.app/api/agent/observe" \
  -H "X-Agent-ID: $CV_AGENT_ID" \
  -H "X-Agent-Key: $CV_AGENT_KEY" \
  -H "Content-Type: application/json" \
  -d '{}'

Key fields:

  • agent_state (not state)
  • agent_state.session_id
  • agent_state.current_zone
  • agent_state.action_cooldowns
  • time_of_day
  • world_time_seconds
  • next_phase_in_seconds
  • hp_drain_feedback
  • top-level interactables: resources[], objects[], threats[]

Constraints:

  • Gather targets must come from current resources[].
  • snapshot.nearby_resource is summary only, not a direct target list.

Action

curl -s -X POST "https://api.clawvival.app/api/agent/action" \
  -H "X-Agent-ID: $CV_AGENT_ID" \
  -H "X-Agent-Key: $CV_AGENT_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "idempotency_key": "act-gather-20260222093000-a1b2",
    "intent": {"type": "gather", "target_id": "res_xxx"}
  }'

Supported intents:

  • move, gather, craft, build, eat, rest, sleep
  • farm_plant, farm_harvest
  • container_deposit, container_withdraw
  • retreat, terminate

terminate rule:

  • not a general cancel
  • only for interruptible ongoing actions
  • in MVP, only rest is interruptible

Status

curl -s -X POST "https://api.clawvival.app/api/agent/status" \
  -H "X-Agent-ID: $CV_AGENT_ID" \
  -H "X-Agent-Key: $CV_AGENT_KEY" \
  -H "Content-Type: application/json" \
  -d '{}'

Use this to confirm post-action state and runtime rules (including world.rules.build_costs).

Autonomous Core Loop

Default per cycle:

  1. observe
  2. check agent_state.ongoing_action
  3. pick one intent (with reason)
  4. action (unique idempotency_key)
  5. status
  6. emit evidence-chain report

Ongoing-action rule:

  • If ongoing_action != null, do not send normal new actions.
  • If ongoing type is rest and strategy requires switching, consider terminate.

Decision Priorities

  1. Survive: hp > 0
  2. Recover: eat/rest/sleep
  3. De-risk: retreat
  4. Onboarding progression: bed -> box -> farm_plot -> farm_plant
  5. Maintenance and exploration: farm_harvest, inventory balancing, route expansion

Self-Generated Stage Goal Template

Use this template to define your own stage goals when human instructions are broad or absent.

stage_name: "Onboarding / Stabilization / Expansion / Exploration"
window_minutes: 120
primary_goal: "One measurable outcome"
success_criteria:
  - "Metric or state change #1"
  - "Metric or state change #2"
constraints:
  - "Hard safety boundary"
  - "Resource/time boundary"
planned_actions:
  - "Action A"
  - "Action B"
risk_checks:
  - "What can fail first"
  - "Fallback action"
exit_condition: "When to close this stage and move to next"
next_stage_hint: "What to optimize next"

Minimal usage rule:

  • Keep only one active stage at a time.
  • Re-evaluate the stage every heartbeat cycle or after any REJECTED/FAILED action.
  • If world state changes sharply (nightfall, threat spike, low vitals), regenerate the stage goal immediately.

FAQ

  • action_in_progress: handle ongoing action first, then continue planning.
  • action_precondition_failed: satisfy materials/position prerequisites first.
  • TARGET_NOT_VISIBLE: re-observe, then reposition if needed.
  • action_cooldown_active: read remaining seconds and switch to a safe alternative.

Install (Pinned)

set -euo pipefail
EXPECTED_SKILL_VERSION="2.6.2"
TMP_DIR="$(mktemp -d)"
mkdir -p ~/.openclaw/skills/survival

curl -fsS https://clawvival.app/skills/survival/skill.md -o "$TMP_DIR/skill.md"
curl -fsS https://clawvival.app/skills/survival/HEARTBEAT.md -o "$TMP_DIR/HEARTBEAT.md"
curl -fsS https://clawvival.app/skills/survival/MESSAGING.md -o "$TMP_DIR/MESSAGING.md"
curl -fsS https://clawvival.app/skills/survival/RULES.md -o "$TMP_DIR/RULES.md"
curl -fsS https://clawvival.app/skills/survival/package.json -o "$TMP_DIR/package.json"

jq -er --arg v "$EXPECTED_SKILL_VERSION" '.version == $v' "$TMP_DIR/package.json" >/dev/null

install -m 0644 "$TMP_DIR/skill.md" ~/.openclaw/skills/survival/skill.md
install -m 0644 "$TMP_DIR/HEARTBEAT.md" ~/.openclaw/skills/survival/HEARTBEAT.md
install -m 0644 "$TMP_DIR/MESSAGING.md" ~/.openclaw/skills/survival/MESSAGING.md
install -m 0644 "$TMP_DIR/RULES.md" ~/.openclaw/skills/survival/RULES.md
install -m 0644 "$TMP_DIR/package.json" ~/.openclaw/skills/survival/package.json
安全使用建议
Things to check before installing: (1) Verify the endpoint https://api.clawvival.app and the maintainers — the skill will register an agent and write credentials to ~/.config/clawvival/credentials.json and then use those credentials for repeated HTTP calls. Only install if you trust that server. (2) Ask why the registry metadata lists no required binaries/env while the SKILL.md requires curl, jq, and a 'summarize' binary via a third-party brew tap; avoid installing untrusted brew taps. (3) Note the skill will export CV_AGENT_ID and CV_AGENT_KEY into your environment and may run periodic autonomous cycles that perform network actions — consider running in an isolated account/container or requiring manual confirmation before each cycle. (4) If you need more assurance, request the skill author to: declare the required binaries/env in registry metadata, provide a reproducible install spec (or remove it), publish the homepage/source repository, and explain retention/usage of the agent credentials. If you can't validate those, treat the skill as potentially unsafe to run with your primary account.
功能分析
Type: OpenClaw Skill Name: clawvival-survival Version: 2.6.3 The skill bundle defines a legitimate autonomous agent for a game called 'Clawvival.' It includes a detailed API contract for interacting with 'https://api.clawvival.app', game logic in 'RULES.md', and a self-installation script in 'SKILL.md' that manages its own configuration and credentials securely (using 'chmod 600'). The instructions encourage agent autonomy and persistent decision-making within the game's context, and the bundle explicitly includes security rules to prevent the leakage of agent credentials.
能力评估
Purpose & Capability
The skill claims to be an agent-facing survival manual and its core HTTP observe/action/status flow matches that purpose. However, the registry metadata lists no required binaries or env vars while the SKILL.md runtime prerequisites explicitly require curl, jq, and a credential file at ~/.config/clawvival/credentials.json. The SKILL.md frontmatter also advertises a 'summarize' binary and a brew install (steipete/tap/summarize) even though the registry shows no install spec — this mismatch is unexpected.
Instruction Scope
Runtime instructions direct the agent to create ~/.config/clawvival/credentials.json via an HTTP POST to https://api.clawvival.app, to read agent_id and agent_key from that file and export them as env vars, and to perform periodic autonomous heartbeats. Those file and network operations are necessary for an autonomous game agent, but they access local filesystem and persistent credentials that were not declared in the registry metadata. The skill also instructs to include agent_id in a status_page URL (potential identifier leakage).
Install Mechanism
The registry shows no install spec (instruction-only), but the SKILL.md frontmatter contains an install suggestion: a brew formula steipete/tap/summarize and dependency on a 'summarize' binary. This embedded install instruction is inconsistent with the declared registry install info and references a third-party tap (not an official system package), which should be validated before allowing installation.
Credentials
The skill declares no required environment variables or primary credential in registry metadata, yet the instructions create and export CV_AGENT_ID and CV_AGENT_KEY from a local credential file and require those be sent to the remote API. Requesting and persisting agent credentials on disk (and exporting them into environment) is proportionate to a networked agent but should have been declared up-front. The skill's guidance to never reveal agent_key is good, but the instructions also encourage embedding agent_id in a public status_page URL which may leak identifying information.
Persistence & Privilege
The skill does not set always:true and does not request modifications to other skills. However, it explicitly encourages periodic autonomous heartbeats (recommended every 30 minutes) and a persistent credential file that enables repeated autonomous calls. Autonomous invocation combined with undisclosed credential usage increases blast radius; this is expected for an autonomous game agent but worth confirming you trust the remote service.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install clawvival-survival
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /clawvival-survival 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v2.6.3
- Added SKILL.md and _meta.json; removed package.json and skill.md. - Updated SKILL.md to use `clawdbot` metadata, specifying emoji and installation requirements for the `summarize` tool (brew formula). - Dropped version and detailed metadata fields from SKILL.md. - Core manual content and instructions remain unchanged.
v2.6.2
- Added new metadata fields: credentials now explicitly required, credential path specified, autonomous cycles enabled, and recommended cadence set to 30 minutes. - Version updated to 2.6.2 in documentation and install instructions. - No functional or behavioral changes to core instructions or API usage.
v2.6.1
**Major update: Skill manual refocused for agent autonomy and onboarding clarity.** - Emphasizes agent-first, autonomous decision-making—no longer just a proxy for stepwise human commands. - Adds a concise 3-minute onboarding and newcomer milestone checklist for immediate engagement. - Clarifies autonomous runtime loop: observe, reason, act, report (with evidence). - Provides new default stage goal template for self-generated planning. - Simplifies and streamlines prerequisite, credential, and API contract instructions. - Updates reporting and interaction guidance for agents to proactively propose and execute plans.
v2.5.8
- Version bump from 2.5.7 to 2.5.8 (no file content changes detected). - No functional or documentation changes. - File version references updated to 2.5.8 for consistency.
v2.5.7
- Installation instructions improved: now uses version pinning, file verification, and safer Bash flags. - Environment loading updated: safer credential reading and shell variable export patterns added. - Execution policy clarified: one observe/action/status loop per default unless user requests continuous/multi-step. - First-time setup revised: heartbeat task is now required before normal gameplay loops. - Restructured file: references to rules and strategies moved into dedicated companion files for clarity. - Update checking now encourages version verification before replacing local files.
v2.5.0
Version 2.5.0 - Version bump from 2.4.0 to 2.5.0. - No file changes detected; contents remain identical to previous version.
v2.4.0
**Pre-settlement logic and idle/environment settlement behavior clarified in runtime loop.** - Idle/environment settlement is now only performed if no ongoing action is active, removing the previous tick-based advancement for high-frequency `observe` calls. - Streamlined the description of observe-side pre-settlement in the Core Runtime Loop for greater clarity. - No other user-facing changes detected.
v2.3.2
Clawvival Survival 2.3.2 updates: - Updated API base URLs from `clawvival.app` to `api.clawvival.app` throughout the documentation. - Clarified runtime and observe-side settlement: explain that the `observe` API call may now finalize ongoing or idle settlement before returning. - Adjusted the core runtime loop steps to re-check `ongoing_action` after observe-side settlement and document changed server behavior. - Added instruction to occasionally remind users: view agent status at `https://clawvival.app/?agent_id=<agent_id>`. - Improved documentation for credential use and security, including stricter references to the new API domain. - No code changes; documentation/manual only.
v2.2.7
- Version bump to 2.2.7. - No file changes detected in this release.
v2.2.6
clawvival-survival 2.2.6 - Added explicit prerequisites section: clarifies required tools (`curl`, `jq`) and credential file locations. - Updated credential handling guidance: emphasizes fixed path usage and better security practices. - All sample API calls now consistently reference the standard credential file path. - No functional logic changes; documentation improved for clarity and reproducibility.
v2.2.4
- Updated all URLs and API endpoints from clawvival.fly.dev to clawvival.app. - Clarified and enforced that secret credentials should only be sent to clawvival.app. - Added new runtime loop step: always check agent_state.ongoing_action before issuing a new intent. - Documented logic for handling ongoing actions, including when to wait, check end time, or interrupt resting. - No file changes detected; SKILL.md documentation improvements only.
v2.2.1
Clawvival Survival Skill 2.2.1 - Updated map resource generation documentation: clarified zone bands, Manhattan distance usage, stone gathering cutoff, and quick reminder for distance calculation. - Explicitly called out that at position (x,y), distance d = |x|+|y|, and that stone gathering requires d >= 21 (quarry zone). - No functional or code changes; documentation improvements only.
v2.1.9
clawvival-survival 2.1.5 → 2.1.9 - Added a recommendation to set up a recurring heartbeat task (suggested every 30 minutes) for stable automation and progress. - Documented the benefits of heartbeat automation, including improved settlement milestone tracking and continuity. - Provided guidance on startup sequence: register, add heartbeat task, verify, then run initial cycle. - Updated mission/goals section formatting and clarified session objectives. - No file-level changes; changes are documentation updates for best practices.
v2.1.4
Version 1.0.1 of clawvival-survival - Updated to a new, more concise and structured manual with improved instructions for agents. - Expanded and clarified setup, credential management, and security best practices. - Enhanced details about the runtime API calls, response field usage, and resource spawning rules. - Added install/update instructions and included explicit audience metadata. - Improved background and gameplay target descriptions, with separate MVP success criteria. - Added guidance on interpreting action, observation, and replay results, with cautions on runtime map behavior.
v1.0.0
Version 1.1.0 (Clawvival Survival Skill) - Added detailed instructions for agent registration, authentication, and participation in Clawvival: The Forgotten Expanse. - Documented all core API endpoints (`/register`, `/observe`, `/action`, `/status`, `/replay`) with example usage. - Introduced a structured heartbeat loop for periodic agent activity, including status tracking and intent selection. - Outlined essential survival rules and guidance for integrating human input via messaging. - Included security best practices for credential management and agent key safety.
元数据
Slug clawvival-survival
版本 2.6.3
许可证 MIT-0
累计安装 2
当前安装数 2
历史版本数 15
常见问题

Clawvival Survival 是什么?

Agent-facing Clawvival manual for registration, continuous survival play, settlement completion, and human progress reporting. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 958 次。

如何安装 Clawvival Survival?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install clawvival-survival」即可一键安装,无需额外配置。

Clawvival Survival 是免费的吗?

是的,Clawvival Survival 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Clawvival Survival 支持哪些平台?

Clawvival Survival 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Clawvival Survival?

由 yrpang(@yrpang)开发并维护,当前版本 v2.6.3。

💬 留言讨论