pctx — MCP Aggregation & Code Mode

MCP aggregation and Code Mode execution layer for token-efficient agent workflows. Wraps portofcontext/pctx — connects agents to Linear, GitHub, and other MC...

What to check before installing: - Confirm platform: the scripts use launchd (macOS). If you're not on macOS, the daemon/start logic won't work. - Review third-party sources: verify the Homebrew tap (portofcontext/tap) and the npm package (@tacticlaunch/mcp-linear) authors/URLs before running install.sh or npm install -g. - Secrets handling: adding MCPs will store API keys in ~/.config/pctx/pctx.json and create timestamped backups there. Decide whether you accept these files (they are chmod 600 but persist on disk); if not, avoid adding upstreams or ensure you have secure backup/cleanup policies. - Persistent daemon: inspect the launchd plist (~/Library/LaunchAgents/ai.openclaw.pctx.plist) before loading it. A running daemon exposes a local HTTP endpoint that can forward requests to external MCPs — understand and control which MCPs/endpoints you connect. - Local code execution: Code Mode runs user-supplied TypeScript in Deno; this enables powerful automation but also can execute arbitrary logic inside the sandbox. Limit which agents/users can send code to the /mcp endpoint. - Discrepancies to clarify with the publisher: SKILL.md frontmatter lists required tools (brew, npm, node) but the registry required binaries is empty, and the skill lacks an official install spec. Ask the maintainer to document supported OSes and to include a safe install manifest or package. If you proceed, run install.sh in a controlled environment, inspect the generated plist and ~/.config/pctx/pctx.json, and avoid adding upstream MCPs until you’ve reviewed the trustworthiness of their tokens and endpoints.

Type: OpenClaw Skill Name: pctx Version: 1.0.1 The pctx skill provides an MCP aggregation layer that manages sensitive API tokens (GitHub, Linear) and executes TypeScript via a local Deno sandbox. While its stated purpose is legitimate, the `pctx-skill.sh` script contains a shell injection vulnerability in the `cmd_test` function. Specifically, the script interpolates user-controlled variables (`ns` and `fn`) directly into a Python command string executed via `python3 -c`, which could allow an attacker or a malicious prompt to execute arbitrary shell commands. Additionally, the skill implements persistence via `launchd` and manages secrets in `~/.config/pctx/pctx.json`, which are high-risk behaviors that, combined with the injection flaw, justify a suspicious classification.