Session Context Injector

Reorient a Telegram chat after a session reset. Reads a project's STATUS.md (resume point, blockers, next action) and sends a project-specific context inject...

Before installing or enabling: 1) Confirm how the skill obtains the bot token — metadata lists TELEGRAM_BOT_TOKEN but SKILL.md shows an `op read` (1Password) step; clarify which method will be used and update metadata. 2) Verify PROJECTS_DIR and the locations of sessions.json / memory/telegram-groups.json — these filesystem reads are not declared and may expose project content; ensure the agent runs in a workspace you trust. 3) Review the Telegram bot's scope and which chats the bot can message (the skill contains a hard-coded special-case chat_id for Nissan). Limit the bot token permissions and rotate it if you enable the skill. 4) Ask the publisher to remove or explain hard-coded IDs, and to declare any CLI tools (e.g., 1Password CLI) needed at runtime. 5) Test in dry-run mode or a sandboxed environment first to ensure messages, truncation, and HTML formatting behave as expected and that no unintended data is leaked. If you cannot get answers to the above, treat the inconsistencies as a risk and do not grant the bot token or filesystem access.

Type: OpenClaw Skill Name: session-context-injector Version: 1.0.0 The skill is designed to read local project status files and exfiltrate their content to the Telegram Bot API (api.telegram.org) to provide context to chat sessions. While this aligns with the stated purpose, it utilizes high-risk capabilities including outbound network access and local file system reads. Specific indicators include the use of a hardcoded Telegram user ID ('821071206') for specialized reporting and a potential path traversal vulnerability in the 'parse_status' function within SKILL.md, where the 'slug' parameter is used to construct file paths without explicit sanitization.