The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill communicates with an external endpoint defined by the $AGENTMEMORY_URL environment variable to fetch commit history. This is a primary function of the skill.
[CREDENTIALS_UNSAFE]: The skill utilizes the environment variable $AGENTMEMORY_SECRET to authenticate its requests. Accessing secrets from environment variables is a standard and recommended practice for agent skills.
[DATA_EXFILTRATION]: The skill ingests commit data from an external source (MCP tool or HTTP endpoint) and renders it into the agent's context. This presents a potential surface for indirect prompt injection if commit messages contain malicious instructions.
Ingestion points: Data returned from the memory_commits MCP tool or the GET request to $AGENTMEMORY_URL (referenced in SKILL.md).
Boundary markers: Absent; the skill renders the commit list directly without explicit delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill only has the capability to render text; it does not include scripts, subprocess calls, or file-system writes.
Sanitization: The skill implements good practices by URL-encoding outgoing parameters (branch, repo, limit) to prevent parameter injection, though it does not specify sanitization for the retrieved commit messages.

commit-history