feat(cli): human-in-the-loop guard for dangerous operations (POC)#159

Open

tonychang04 wants to merge 15 commits into

mainfrom

feat/hitl-guard-poc

Commits on Jun 10, 2026

feat(cli): human-in-the-loop guard for dangerous operations (POC)

tonychang04
and
claude
committed
refactor(cli): agent supplies the explanation, CLI makes no LLM call

tonychang04
and
claude
committed
feat(cli): nudge the calling agent to articulate intent + render its brief

claude
authored and
tonychang04
committed
feat(cli): tailor guard facts to the live project

claude
authored and
tonychang04
committed
feat(cli): add 'what this means for your users' to the guard page

claude
authored and
tonychang04
committed
style(cli): match the guard approval page to the InsForge brand

claude
authored and
tonychang04
committed
style(cli): redesign the approval page, drop the 'Verified by InsForge' dividers

claude
authored and
tonychang04
committed
fix(cli): use npx @insforge/cli in echoed command + fix approval button colors

claude
authored and
tonychang04
committed
feat(cli): make the guard approval window configurable

claude
authored and
tonychang04
committed
test(cli): cover the guard's danger verdict and brief assembly

claude
authored and
tonychang04
committed
feat(cli): let the agent escalate the guard (escalate-only) for edge cases

claude
authored and
tonychang04
committed
feat(cli): gate the guard behind INSFORGE_GUARD (default off) for safe rollout

claude
authored and
tonychang04
committed
feat(cli): persist guard enablement via 'link --guard'

claude
authored and
tonychang04
committed
fix(cli): address PR review — guard bypass, CSRF, and correctness

claude
authored and
tonychang04
committed
fix(cli): flip the SQL gate from denylist to allowlist (fail-closed)

claude
authored and
tonychang04
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(cli): human-in-the-loop guard for dangerous operations (POC)#159

feat(cli): human-in-the-loop guard for dangerous operations (POC)#159
tonychang04 wants to merge 15 commits into
mainfrom
feat/hitl-guard-poc

Commits on Jun 10, 2026

feat(cli): human-in-the-loop guard for dangerous operations (POC)

refactor(cli): agent supplies the explanation, CLI makes no LLM call

feat(cli): nudge the calling agent to articulate intent + render its brief

feat(cli): tailor guard facts to the live project

feat(cli): add 'what this means for your users' to the guard page

style(cli): match the guard approval page to the InsForge brand

style(cli): redesign the approval page, drop the 'Verified by InsForge' dividers

fix(cli): use npx @insforge/cli in echoed command + fix approval button colors

feat(cli): make the guard approval window configurable

test(cli): cover the guard's danger verdict and brief assembly

feat(cli): let the agent escalate the guard (escalate-only) for edge cases

feat(cli): gate the guard behind INSFORGE_GUARD (default off) for safe rollout

feat(cli): persist guard enablement via 'link --guard'

fix(cli): address PR review — guard bypass, CSRF, and correctness

fix(cli): flip the SQL gate from denylist to allowlist (fail-closed)