The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill directs the agent to generate and execute shell commands through direct string interpolation of user-supplied variables (such as book titles and character names). This pattern is highly susceptible to command injection attacks, as malicious users could provide inputs containing shell metacharacters like $(command) or backticks to execute arbitrary code.\n- [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to initialize project structures and run Python scripts (e.g., webnovel.py). These bash commands are constructed using templates that do not safely handle user input, leading to potential shell subversion.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content via WebFetch and user-provided reference book excerpts. This data is passed to sub-agents and used to populate configuration files, creating a vulnerability surface where malicious instructions in the source text could hijack the agent's logic.\n
Ingestion points: AskUserQuestion (collecting灵感来源 and book excerpts), WebFetch (fetching market trends).\n
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat external content as untrusted data.\n
Capability inventory: The skill has access to shell execution (Bash), file writing (Write, Edit), and agent delegation (Agent), which can be abused if a prompt injection occurs.\n
Sanitization: Although the skill provides a Python regex for sanitizing project slugs, this logic is itself executed within a bash command that is vulnerable to interpolation-based injection.

webnovel-init