The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several standard command-line tools such as git, npm, gh, rg, and pnpm to perform repository management tasks, including branch creation, version verification, documentation builds, and pull request creation.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from apps/kimi-code/CHANGELOG.md, which creates a surface for indirect prompt injection if malicious instructions were merged into the upstream source. The risk is minimized by the skill's specific rules for data transformation.
Ingestion points: apps/kimi-code/CHANGELOG.md is read to extract version blocks and entry body text.
Boundary markers: The instructions provide rigid logic for parsing, stripping decorations, and classifying entries, which helps maintain a clear separation between content and instructions.
Capability inventory: The agent possesses capabilities for file system writes to documentation paths, shell command execution for repository operations, and automated pull request generation.
Sanitization: Mandatory instructions to remove PR links, commit hashes, and 'Thanks' credits, along with requirements to replace internal identifiers with neutral placeholders, serve as effective sanitization layers.

sync-changelog