Lambda Lang
PassAudited by ClawScan on May 1, 2026.
Overview
Lambda Lang appears to be a coherent language/reference skill, but users should not let compressed agent-to-agent messages become trusted commands without review.
This skill is reasonable to install as a Lambda language reference/parser aid. Before using it operationally, make sure your agent decodes Lambda messages, confirms sensitive commands with you, authenticates any peer agents, and avoids saving new vocabulary or context definitions from untrusted sources.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A peer or prompt could encode an instruction in Lambda that looks compact or opaque to a human.
Lambda intentionally supports compact command and task-delegation messages between agents. This is central to the skill, but it means Lambda text from another agent could contain action requests.
**Types**: `?` query · `!` assert · `.` command ... **When to use Lambda** - Agent-to-agent messaging ... task delegation
Treat Lambda messages as untrusted input until decoded, and require normal user approval before executing sensitive or high-impact actions.
A bad or confusing definition could cause later Lambda messages to be interpreted incorrectly.
The spec allows conversation-local terms and promotion into shared vocabulary, which can shape how later compact messages are interpreted.
Agents can define new terms mid-conversation: `{def:xyz="complex concept description"}` ... Definitions are scoped to conversation unless promoted to shared vocabulary.Do not persist or promote new vocabulary definitions from untrusted conversations without review; keep definitions scoped to the current task when possible.
If used with external agents, Lambda messages may cross trust boundaries and should not be assumed to come from a trusted peer.
The documentation shows optional use of Lambda over an external agent messaging protocol, including send and publish flows.
pilotctl send-message target-agent --data "?Uk/co" --type text ... pilotctl publish target-agent "lambda:co" --data "!co.ex"
Authenticate peer agents, log or decode Lambda content for human review when needed, and avoid acting on messages from unknown agents.