skill-c-fetch-minutes

【会后纪要抓取与issue草稿生成】每10分钟由 OpenClaw cron 触发一次。负责在会议结束后从腾讯会议拉取转录与AI智能纪要，由 OpenClaw 做两阶段issue抽取，生成 draft_issue.md 等四个文件并提交到 Gitea，最后通知组织者审核。不处理会议创建、会前简报、issue落地等场景。

Before installing or enabling this skill: - Note the manifest/registry declares no required env vars, but the code and setup script require a config file at ~/.config/skill-c-fetch-minutes/.env and the variables GITEA_BASE_URL, GITEA_TOKEN_BOT, and AIFUSION_META_REPO. Do not provide credentials until you confirm these requirements. - The GITEA_TOKEN_BOT grants repository and meta-repo access (read/write). Use a dedicated bot account with the minimum necessary scopes (only the specific repos and meta-repo), not a personal or org-wide admin token. - The env-example contains an IP address (43.156.243.152:3000). Verify the intended Gitea host and do not point credentials to unknown hosts. - setup.sh will run pip install -r requirements.txt (uses --break-system-packages). Run setup inside an isolated environment (virtualenv, container) rather than on your system Python to avoid system-wide changes. - Review the code paths that call the Gitea API (gitea_utils.py, log_utils.py, commit_content.py, scan.py) to ensure behaviour and targets are acceptable (which repos will be enumerated via /repos/search and what the token can access). - If you want to proceed, consider: create a least-privilege bot token limited to specific repos and meta repo; run the skill in an isolated environment; and test against a staging Gitea instance first. If you want, I can list the exact lines where the code reads env vars and calls Gitea APIs and suggest minimal token scopes to grant.

Type: OpenClaw Skill Name: skill-c-fetch-minutes Version: 1.0.2 The skill bundle automates meeting minutes processing but contains a significant security vulnerability in `scripts/commit_content.py`. The script accepts arbitrary file paths via command-line arguments (e.g., `--transcript-file`) and uploads their contents to Gitea without path sanitization or directory restricted validation. This creates a risk of arbitrary file exfiltration; if the AI agent is manipulated via prompt injection to provide paths to sensitive local files (such as `~/.ssh/id_rsa` or `.env` files), the script will read and upload them to the Gitea repository. While the behavior aligns with the stated purpose, the lack of input validation on file operations is a high-risk flaw.