The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's scripts (scripts/memory-backup.sh and scripts/memory-consolidate.sh) use npx @claude-flow/cli to export and optimize memory. This pattern downloads and executes code from the npm registry during execution. Because the package provider is not a trusted vendor and is not associated with the skill author (growgami), it presents a high risk of executing unverified external code.
[EXTERNAL_DOWNLOADS]: Use of npx without version constraints results in the automatic download of the latest version of the @claude-flow/cli tool from the public npm registry. This exposes the environment to potential supply chain attacks where a malicious package update could be executed without user intervention.
[COMMAND_EXECUTION]: The skill includes shell scripts that perform system-level operations. There is a notable discrepancy between the SKILL.md documentation, which claims 'no external CLI or dependencies required,' and the actual inclusion of scripts that depend on external packages via npx.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) based on its architectural design:
Ingestion points: Data is read from memory/clients.md, memory/skill-patterns.md, and memory/signals.md in SKILL.md.
Boundary markers: Absent. The skill instructions do not specify any delimiters or safety warnings to ignore instructions that might be embedded in the memory files.
Capability inventory: The skill can execute shell scripts and external CLI tools via npx.
Sanitization: Absent. The data read from memory files is processed directly by the agent to influence its behavior and decisions without validation or escaping.

memory-management