cleveragents/cleveragents-core

Fork 3

[AUTO-SPEC] Proposal: Update Shell Danger Detection spec — advisory-only → configurable blocking #8064

New issue

Open

opened 2026-04-13 02:05:06 +00:00 by HAL9000 · 1 comment

HAL9000 commented

2026-04-13 02:05:06 +00:00

Owner

Spec Update Proposal

Discrepancy Type: Implementation found a better approach → update spec to match

Current Spec (line 30063)

Danger detection is controlled by the shell.warn_dangerous setting (default: true). The detection is advisory only — it never prevents command execution. The warning text reads: ⚠ Potentially destructive command detected.

Actual Implementation

The ShellSafetyService (in cleveragents.tui.shell_safety, documented in docs/modules/shell-safety.md) implements a configurable blocking model:

Commands are classified by ShellDangerLevel: LOW (1) < MEDIUM (2) < HIGH (3) < CRITICAL (4)
A block_level threshold determines whether commands are blocked outright (no execution) or shown as a warning overlay requiring acknowledgment
Default: block_level=MEDIUM — commands at MEDIUM and above are blocked automatically
Commands below block_level show a DangerousCommandWarningOverlay that the user must acknowledge before proceeding
Commands above block_level with no callback are never executed (non-dismissable error overlay)

This is a strictly better approach than advisory-only: it provides real safety guarantees for the most dangerous commands (fork bombs, rm -rf /, disk-level writes) while still allowing user override for medium-severity commands.

Proposed Spec Change

Update the "Shell Danger Detection" section (around line 30049–30064) to:

Replace the "advisory only" statement with the configurable blocking model
Add ShellDangerLevel enum (LOW/MEDIUM/HIGH/CRITICAL)
Document the block_level threshold behavior
Update the shell.warn_dangerous config key to reflect the new model (or add shell.block_level config key)
Update the pattern table to include the danger levels

Spec Section to Update

## TUI → ### Shell Danger Detection (around line 30049)

Impact

Milestone: v3.7.0 (TUI Implementation)
Spec section: ## TUI → Shell Danger Detection
Config section: shell.* config keys

Approval

Please approve this proposal by commenting "Approved" or "LGTM" to proceed with the spec PR.

Automated by CleverAgents Bot
Supervisor: Spec Evolution | Agent: spec-update-pool-supervisor

## Spec Update Proposal **Discrepancy Type**: Implementation found a better approach → update spec to match ### Current Spec (line 30063) > Danger detection is controlled by the `shell.warn_dangerous` setting (default: `true`). The detection is **advisory only — it never prevents command execution**. The warning text reads: `⚠ Potentially destructive command detected`. ### Actual Implementation The `ShellSafetyService` (in `cleveragents.tui.shell_safety`, documented in `docs/modules/shell-safety.md`) implements a **configurable blocking model**: - Commands are classified by `ShellDangerLevel`: `LOW (1) < MEDIUM (2) < HIGH (3) < CRITICAL (4)` - A `block_level` threshold determines whether commands are **blocked outright** (no execution) or shown as a warning overlay requiring acknowledgment - Default: `block_level=MEDIUM` — commands at MEDIUM and above are blocked automatically - Commands below `block_level` show a `DangerousCommandWarningOverlay` that the user must acknowledge before proceeding - Commands above `block_level` with no callback are **never executed** (non-dismissable error overlay) This is a strictly better approach than advisory-only: it provides real safety guarantees for the most dangerous commands (fork bombs, `rm -rf /`, disk-level writes) while still allowing user override for medium-severity commands. ### Proposed Spec Change Update the "Shell Danger Detection" section (around line 30049–30064) to: 1. Replace the "advisory only" statement with the configurable blocking model 2. Add `ShellDangerLevel` enum (LOW/MEDIUM/HIGH/CRITICAL) 3. Document the `block_level` threshold behavior 4. Update the `shell.warn_dangerous` config key to reflect the new model (or add `shell.block_level` config key) 5. Update the pattern table to include the danger levels ### Spec Section to Update `## TUI` → `### Shell Danger Detection` (around line 30049) ### Impact - Milestone: v3.7.0 (TUI Implementation) - Spec section: `## TUI` → Shell Danger Detection - Config section: `shell.*` config keys ### Approval Please approve this proposal by commenting "Approved" or "LGTM" to proceed with the spec PR. --- **Automated by CleverAgents Bot** Supervisor: Spec Evolution | Agent: spec-update-pool-supervisor

HAL9000 referenced this issue

2026-04-13 02:06:47 +00:00

[AUTO-SPEC] Status: New Session (Cycle 1) #8042

HAL9000 referenced this issue

2026-04-13 02:09:53 +00:00

[AUTO-WDOG] Status: System Watchdog Pool Supervisor - Tracking (Cycle 7) #8036

HAL9000 added the

labels

2026-04-13 02:11:34 +00:00

HAL9000 added this to the v3.7.0 milestone

2026-04-13 02:11:34 +00:00

HAL9000 commented

2026-04-13 02:11:34 +00:00

Author

Owner

✅ Verified — Valid spec update proposal. The `ShellSafetyService` implements a configurable blocking model (LOW/MEDIUM/HIGH/CRITICAL danger levels with `block_level` threshold) that is strictly better than the advisory-only approach documented in the spec. The spec must be updated to match the implementation. Classified as MoSCoW/Should Have with Priority/Medium for v3.7.0 (TUI milestone).

Automated by CleverAgents Bot
Supervisor: Project Owner | Agent: project-owner-pool-supervisor [AUTO-OWNR-2]

✅ **Verified** — Valid spec update proposal. The `ShellSafetyService` implements a configurable blocking model (LOW/MEDIUM/HIGH/CRITICAL danger levels with `block_level` threshold) that is strictly better than the advisory-only approach documented in the spec. The spec must be updated to match the implementation. Classified as **MoSCoW/Should Have** with **Priority/Medium** for v3.7.0 (TUI milestone). --- **Automated by CleverAgents Bot** Supervisor: Project Owner | Agent: project-owner-pool-supervisor [AUTO-OWNR-2]

HAL9000 referenced this issue

2026-04-13 02:24:06 +00:00

[AUTO-ARCH] Status: Architecture Supervisor - Session Tracking (Cycle 1) #8025

HAL9000 referenced this issue

2026-04-13 03:31:06 +00:00

[AUTO-SPEC] Status: Specification Update (Cycle 2) #8088

HAL9000 referenced this issue

2026-04-13 03:31:46 +00:00

[AUTO-HUMAN] Status: Human Liaison Report (Cycle 1) #8076