The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/install.sh script performs a git clone of the revfactory/webtoon-harness repository from GitHub to obtain the 27 agent definitions and methodology skills required for the harness.
[COMMAND_EXECUTION]: The skill instructs the user to execute bash scripts/install.sh, which performs network operations (Git) and file system modifications to scaffold the harness into the project's .claude/ directory.
[DYNAMIC_EXECUTION]: The harness dynamically loads and executes 27 specialized agents and multiple skills from the downloaded files using Claude Code's agent-team functionality. This involves executing instructions defined in external markdown and script files.
[INDIRECT_PROMPT_INJECTION]: The research phase (Phase 2) uses the scrapling skill to ingest untrusted data from web platforms (trends, rankings, and reader comments). This data is used to generate the planning brief that drives the subsequent scenario and script generation phases.
Ingestion points: Web trend data extracted via scrapling in references/trend-research-scrapling.md and SKILL.md.
Boundary markers: Absent; while the skill emphasizes recording source URLs and respecting terms of service, it does not specify the use of delimiters or 'ignore' instructions for the agent when processing the retrieved web content.
Capability inventory: The skill has access to powerful tools including Bash for command execution, the Agent tool for multi-agent coordination, and the codex CLI for image generation.
Sanitization: The analysis did not find evidence of sanitization or validation of the ingested web text before it is interpolated into prompts for the scenario generation team.

webtoon-harness