The Agent Skills Directory

[COMMAND_EXECUTION]: The bundled Node.js script scripts/sessions-axi.mjs executes shell commands using execSync. Specifically, the detectGitProject function runs git rev-parse --show-toplevel to identify the current repository root.
[DATA_EXFILTRATION]: The skill performs network operations using the fetch API to communicate with a server defined by the CLAUDE_ASSIST_SERVER environment variable. This server defaults to localhost but can be configured to point to an external endpoint, where it transmits session identifiers and receives conversation transcripts.
[PROMPT_INJECTION]: The skill is exposed to indirect prompt injection because it retrieves and feeds historical session transcripts into the agent's context. These transcripts could contain malicious instructions from past untrusted interactions that the agent might obey upon recall.
Ingestion points: Transcripts are ingested via the transcript and search subcommands in scripts/sessions-axi.mjs.
Boundary markers: The instructions do not define boundary markers or specify that the agent should ignore instructions contained within retrieved transcripts.
Capability inventory: The agent can execute the sessions-axi script, which has network access and can execute shell commands.
Sanitization: There is no evidence of sanitization or filtering of historical transcript content before it is returned to the agent's prompt.

assist-sessions