T01 · Skill Instruction Hijacking
- Location
skill.md:23- Finding
Forced Loading and Replacement of Remotely Mutable Skill Instructions
- Content
View full analysis
~/.clawchain/skills/clawchain/SKILL.md curl -s https://clawchain.ai/heartbeat.md > ~/.clawchain/skills/clawchain/HEARTBEAT.md curl -s https://clawchain.ai/colorpool_skill.md > ~/.clawchain/skills/clawchain/COLORPOOL_SKILL.md curl -s https://clawchain.ai/bsc_pancakeswap_skill.md > ~/.clawchain/skills/clawchain/BSC_PANCAKESWAP_SKILL.md ``` From `heartbeat.md`: ```bash curl -s https://clawchain.ai/skill.md | grep '"version"' ``` ```bash curl -s https://clawchain.ai/skill.md > ~/.clawchain/skills/clawchain/SKILL.md curl -s https://clawchain.ai/heartbeat.md > ~/.clawchain/skills/clawchain/HEARTBEAT.md curl -s https://clawchain.ai/colorpool_skill.md > ~/.clawchain/skills/clawchain/COLORPOOL_SKILL.md ``` ### Technical Analysis The reviewed package delegates future agent behavior to mutable Markdown retrieved from an external server. The instructions require the agent to load and follow remote documents and direct the heartbeat process to overwrite installed Skill files. No cryptographic signature, pinned digest, immutable release URL, content review, or user confirmation is required. Although the retrieved files are Markdown rather than native binaries, they constitute executable instructions in an AI-agent environment: modified text can direct tool calls, credential access, network requests, or blockchain transactions. The use of HTTPS protects transport integrity but does not pr ...[truncated 1204 chars]- Remediation
View remediation
