The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill describes the capability of the atmos init command to fetch project templates from various remote sources.
Evidence includes support for template sources via Git, S3 buckets, OCI registries, and HTTPS URLs.
[COMMAND_EXECUTION]: The tool documented in this skill supports the execution of lifecycle hooks during the project bootstrapping process.
These hooks are defined in the template manifest under spec.hooks and support the execution of sequences of steps.
[DYNAMIC_EXECUTION]: The skill documents the use of the Common Expression Language (CEL) within templates to handle conditional logic for fields and files.
The when: predicates in template manifests allow for dynamic evaluation of expressions over user-provided answers at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface where the agent processes external, potentially untrusted data that can influence its behavior.
Ingestion points: Remote project templates fetched via Git, S3, OCI, or HTTPS as described in SKILL.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are documented for the template ingestion process.
Capability inventory: The atmos CLI, as used by the agent, can perform file system writes and execute shell commands through the spec.hooks mechanism.
Sanitization: No explicit sanitization or validation of the remote template content is mentioned; the security of the operation depends on the user's trust in the template source.

atmos-init