Problem
AGENT-PERMISSIONS.md correctly calls out that the skill's approval guidance is behavior rather than a security boundary, and its Tier 3 section recommends a pre-execution hook for destructive temporal / tcld commands. The current hand-written regex example is intentionally described as a speed bump because aliases, shell indirection, and wrappers can bypass text matching.
Would you be open to adding an optional maintained HOL Guard path alongside that Tier 3 example?
Proposed scope
Keep Temporal's existing credential scoping, namespace delete protection, native confirmations, audit logs, and blast-radius guidance authoritative. Add a short optional section that directly installs and invokes HOL Guard on a supported local coding-agent harness before state-changing Temporal work, for example:
pipx install hol-guard
hol-guard detect --json
hol-guard install <detected-harness>
hol-guard run <detected-harness>
The accepted guidance should be explicit about the boundary:
- HOL Guard protects the supported local agent harness; it does not run inside Temporal Cloud or the Temporal server.
- Temporal credentials, server-side permissions/delete protection, and native CLI safety checks remain authoritative.
- Do not claim dedicated classification for every
temporal / tcld command unless that behavior is independently verified.
- Guard deny/review/error/unavailable states must not be bypassed by relaunching an unprotected agent for the same mutation.
- The existing handcrafted hook can remain as the transparent Temporal-specific example; HOL Guard is the optional reusable runtime boundary for users who already run supported coding agents.
If this placement fits, I can follow with the smallest repository-native docs/skill change and whatever validation you prefer.
HOL Guard: https://github.com/hashgraph-online/hol-guard
Affiliation: I maintain HOL Guard / Hashgraph Online. AI assistance was used to prepare this issue; I checked the proposed boundary against the current Temporal Ops Skill documentation and the maintained HOL Guard setup flow.
Problem
AGENT-PERMISSIONS.mdcorrectly calls out that the skill's approval guidance is behavior rather than a security boundary, and its Tier 3 section recommends a pre-execution hook for destructivetemporal/tcldcommands. The current hand-written regex example is intentionally described as a speed bump because aliases, shell indirection, and wrappers can bypass text matching.Would you be open to adding an optional maintained HOL Guard path alongside that Tier 3 example?
Proposed scope
Keep Temporal's existing credential scoping, namespace delete protection, native confirmations, audit logs, and blast-radius guidance authoritative. Add a short optional section that directly installs and invokes HOL Guard on a supported local coding-agent harness before state-changing Temporal work, for example:
The accepted guidance should be explicit about the boundary:
temporal/tcldcommand unless that behavior is independently verified.If this placement fits, I can follow with the smallest repository-native docs/skill change and whatever validation you prefer.
HOL Guard: https://github.com/hashgraph-online/hol-guard
Affiliation: I maintain HOL Guard / Hashgraph Online. AI assistance was used to prepare this issue; I checked the proposed boundary against the current Temporal Ops Skill documentation and the maintained HOL Guard setup flow.