The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute code from a remote, untrusted GitHub repository via shell piping.
Evidence: curl -fsSL https://raw.githubusercontent.com/nexu-io/looper/main/scripts/install.sh | sh in SKILL.md.
Evidence: curl -fsSL https://raw.githubusercontent.com/nexu-io/looper/main/scripts/uninstall.sh | sh in references/cli.md.
[EXTERNAL_DOWNLOADS]: Fetches installation and maintenance scripts from an external source not belonging to a trusted vendor.
Evidence: Source URL https://raw.githubusercontent.com/nexu-io/looper/main/scripts/install.sh.
[PERSISTENCE]: Configures background processes and system services to ensure the tool remains active across sessions and reboots.
Evidence: looper daemon start --daemon-mode launchd in references/daemon.md to establish a persistent macOS LaunchAgent.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: GitHub issues, pull requests, and Plane work-items identified in SKILL.md and references/plane.md.
Boundary markers: The instructions do not define clear delimiters or warnings to separate processed data from agent instructions.
Capability inventory: High-privilege actions including shell command execution via agent CLIs, file system writes, and daemon lifecycle management.
Sanitization: While the skill suggests redacting secrets in user-facing summaries, it provides no mechanisms for sanitizing the untrusted data ingested for automation loops.
[COMMAND_EXECUTION]: Executes a variety of third-party command-line interfaces and local scripts.
Evidence: Direct execution of agent binaries including claude, grok, pi, omp, git, and gh.
Evidence: Execution of a local diagnostic script via bash <skill-bundle>/scripts/check.sh.
[DYNAMIC_EXECUTION]: Executes scripts from paths that are computed dynamically at runtime.
Evidence: bash <skill-bundle>/scripts/check.sh where the script path is relative to the dynamic installation location of the skill bundle.

looper-looper