pdurlej/platform

Fork 0

fix(platformctl): validate health container inputs #671

Merged

pdurlej merged 1 commit from codex/211-212-health-input-hardening into main

2026-06-01 16:14:16 +02:00

codex commented

2026-06-01 15:25:31 +02:00

Collaborator

Canary status: missing — fire canary 3+3 manually before merge

Canary Context Pack

Product story

platformctl health feeds operator and agent confidence. It must not turn malformed runbook input into a shell command, and hostile health inputs should fail in bounded, readable ways instead of crashing or leaking local paths.

What changed

Added safe container-name validation for spec.runtime.container_name, runbook-derived container names, and default names before docker inspect command construction.
Made malformed or corrupted runbook.md container fields fail closed with no transport call.
Added module-id length and character bounds for health/smoke paths.
Added bounded manifest error summaries for health reports.
Added adversarial tests for shell metacharacters, malformed runbook fields, corrupted runbooks, null bytes, Unicode confusables, very long module IDs, and invalid manifest YAML.

Why it changed

M06 health hardening needs adversarial input coverage so agent-visible health state cannot become false-green or execute unexpected command text derived from repo docs.

Files touched

control-plane/platformctl/health.py
control-plane/platformctl/tests/test_health_phase3.py

Relevant context

M06 apply-hardening closeout
Issue #211: validate runbook-derived container names
Issue #212: add adversarial health input cases
Existing platformctl health Phase 3 behavior

Runtime evidence

No runtime health calls, no SSH, no production mutation, no live DB, no service restart. Tests use fake transports and local fixtures only.

Known constraints

Container name validation intentionally allows Docker-style ASCII names with letters, digits, underscore, dot, and dash only. Directory/file details in manifest failures are intentionally compressed for user-facing health output.

Explicit out-of-scope

No runtime apply.
No break-glass or actor identity changes.
No Forgejo workflow contract parsing.
No changes to unrelated Infisical inventory files currently dirty in the worktree.

Requested decision

Review and merge if checks stay green.

Merge blockers

Any valid current module/runbook container name rejected by the new pattern.
Any path where malformed runbook input still reaches docker inspect.
Any health regression in targeted tests.

Spec sources read

control-plane/platformctl/health.py — health command, runbook-derived container parsing, smoke input handling.
control-plane/platformctl/tests/test_health_phase3.py — existing health tests and adversarial coverage location.
Forgejo issue #211 — runbook-derived container-name validation acceptance criteria.
Forgejo issue #212 — adversarial health input acceptance criteria.

Validation

UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run pytest platformctl/tests/test_health_phase3.py — 27 passed
UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run pytest platformctl/tests/test_apply_phase3.py platformctl/tests/test_apply_env_file.py platformctl/tests/test_pr_sanity.py platformctl/tests/test_forgejo_ci_scripts_contract.py platformctl/tests/test_health_phase3.py — 208 passed
PYTHONPATH=control-plane UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run --project control-plane python -m platformctl.cli validate all --json — passed, exitCode=0

Closes #211
Closes #212

Canary status: missing — fire canary 3+3 manually before merge ## Canary Context Pack ### Product story `platformctl health` feeds operator and agent confidence. It must not turn malformed runbook input into a shell command, and hostile health inputs should fail in bounded, readable ways instead of crashing or leaking local paths. ### What changed - Added safe container-name validation for `spec.runtime.container_name`, runbook-derived container names, and default names before `docker inspect` command construction. - Made malformed or corrupted `runbook.md` container fields fail closed with no transport call. - Added module-id length and character bounds for health/smoke paths. - Added bounded manifest error summaries for health reports. - Added adversarial tests for shell metacharacters, malformed runbook fields, corrupted runbooks, null bytes, Unicode confusables, very long module IDs, and invalid manifest YAML. ### Why it changed M06 health hardening needs adversarial input coverage so agent-visible health state cannot become false-green or execute unexpected command text derived from repo docs. ### Files touched - `control-plane/platformctl/health.py` - `control-plane/platformctl/tests/test_health_phase3.py` ### Relevant context - M06 apply-hardening closeout - Issue #211: validate runbook-derived container names - Issue #212: add adversarial health input cases - Existing `platformctl health` Phase 3 behavior ### Runtime evidence No runtime health calls, no SSH, no production mutation, no live DB, no service restart. Tests use fake transports and local fixtures only. ### Known constraints Container name validation intentionally allows Docker-style ASCII names with letters, digits, underscore, dot, and dash only. Directory/file details in manifest failures are intentionally compressed for user-facing health output. ### Explicit out-of-scope - No runtime apply. - No break-glass or actor identity changes. - No Forgejo workflow contract parsing. - No changes to unrelated Infisical inventory files currently dirty in the worktree. ### Requested decision Review and merge if checks stay green. ### Merge blockers - Any valid current module/runbook container name rejected by the new pattern. - Any path where malformed runbook input still reaches `docker inspect`. - Any health regression in targeted tests. ## Spec sources read - `control-plane/platformctl/health.py` — health command, runbook-derived container parsing, smoke input handling. - `control-plane/platformctl/tests/test_health_phase3.py` — existing health tests and adversarial coverage location. - Forgejo issue #211 — runbook-derived container-name validation acceptance criteria. - Forgejo issue #212 — adversarial health input acceptance criteria. ## Validation - `UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run pytest platformctl/tests/test_health_phase3.py` — 27 passed - `UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run pytest platformctl/tests/test_apply_phase3.py platformctl/tests/test_apply_env_file.py platformctl/tests/test_pr_sanity.py platformctl/tests/test_forgejo_ci_scripts_contract.py platformctl/tests/test_health_phase3.py` — 208 passed - `PYTHONPATH=control-plane UV_CACHE_DIR=/private/tmp/codex-uv-cache uv run --project control-plane python -m platformctl.cli validate all --json` — passed, `exitCode=0` Closes #211 Closes #212

codex added 1 commit

2026-06-01 15:25:31 +02:00

fix(platformctl): validate health container inputs

python-ci / Python 3.12 (pull_request) Successful in 36s

Details

canary-required / collect-diff (pull_request) Successful in 4s

Details

platformctl plan / auto-apply scope (pull_request) Successful in 16s

Details

pyfallow / Pyfallow gate (control-plane) (pull_request) Successful in 15s

Details

python-ci / Python 3.11 (pull_request) Successful in 35s

Details

python-ci / Python 3.13 (pull_request) Successful in 36s

Details

base-is-main / guard (pull_request) Successful in 1s

Details

patchwarden-client-dry-run / collect-diff (pull_request) Successful in 3s

Details

canary-required / canary (pull_request) Successful in 12s

Details

patchwarden-pr-sanity / collect-diff (pull_request) Successful in 4s

Details

patchwarden-client-dry-run / dry-run (pull_request) Successful in 17s

Details

patchwarden-pr-sanity / sanity (pull_request) Successful in 6m11s

Details

6abdc384c2

codex added the

agent/codex

class/security-sensitive

risk/process

tier/full

labels

2026-06-01 15:25:57 +02:00

codex added this to the 06 - Agent execution and CI governance milestone

2026-06-01 15:25:58 +02:00

codex commented

2026-06-01 15:36:15 +02:00

Author

Collaborator

Patchwarden PR sanity

Status: advisory_findings
PR: 671
Commit: 6abdc384c2cffa9e6e9faf245eef373c7f0b21d4
Security-sensitive label: present
Authority: advisory model review plus deterministic blockers only
3+3 canary: still alive; this does not replace it

Deterministic findings

No deterministic findings.

Model reviewers

`global-glm` / `glm-5.1:cloud`

Status: error
Verdict: -
Note: ReadTimeout: The read operation timed out
Findings: none

`global-deepseek` / `deepseek-v4-pro:cloud`

Status: ok
Verdict: NOT_OK
high Runbook Container field parsing change may break existing modules
- Evidence: control-plane/platformctl/health.py: _runbook_container_name now returns an error if any line contains 'Container:' but does not match the exact Container: `name` pattern, causing health_container_name to raise HealthConfigError and con
- Next: Audit all existing runbook.md files for 'Container:' lines that are not valid container names. Consider making the pattern more lenient (e.g., only validate if backticks are present) or emit a warning instead of failing the health check, to avoid breaking modules with documentation-only Container re
medium Module ID length limit may reject existing long module IDs
- Evidence: control-plane/platformctl/health.py: _MODULE_ID_MAX_LENGTH set to 120 and enforced in _validate_module_id, which is called by run_smoke and health_module. Any module ID longer than 120 characters will now fail health checks.
- Next: Verify that no existing module IDs exceed 120 characters. If any do, either increase the limit or rename the modules before merging.

`redteam` / `kimi-k2.6:cloud`

Status: ok
Verdict: OK
Findings: none

Policy notes

GLM 5.1 + DeepSeek V4 Pro are the operator-required model mix for this bot.
Optional red-team model is enabled only when PLATFORMCTL_PR_SANITY_REDTEAM_MODEL is configured.
Auto-merge is not enabled here.

# Patchwarden PR sanity - Status: `advisory_findings` - PR: `671` - Commit: `6abdc384c2cffa9e6e9faf245eef373c7f0b21d4` - Security-sensitive label: `present` - Authority: advisory model review plus deterministic blockers only - 3+3 canary: still alive; this does not replace it ## Deterministic findings No deterministic findings. ## Model reviewers ### `global-glm` / `glm-5.1:cloud` - Status: `error` - Verdict: `-` - Note: ReadTimeout: The read operation timed out - Findings: none ### `global-deepseek` / `deepseek-v4-pro:cloud` - Status: `ok` - Verdict: `NOT_OK` - **`high`** Runbook Container field parsing change may break existing modules - Evidence: `control-plane/platformctl/health.py: _runbook_container_name now returns an error if any line contains 'Container:' but does not match the exact `Container: \`name\`` pattern, causing health_container_name to raise HealthConfigError and con` - Next: Audit all existing runbook.md files for 'Container:' lines that are not valid container names. Consider making the pattern more lenient (e.g., only validate if backticks are present) or emit a warning instead of failing the health check, to avoid breaking modules with documentation-only Container re - **`medium`** Module ID length limit may reject existing long module IDs - Evidence: `control-plane/platformctl/health.py: _MODULE_ID_MAX_LENGTH set to 120 and enforced in _validate_module_id, which is called by run_smoke and health_module. Any module ID longer than 120 characters will now fail health checks.` - Next: Verify that no existing module IDs exceed 120 characters. If any do, either increase the limit or rename the modules before merging. ### `redteam` / `kimi-k2.6:cloud` - Status: `ok` - Verdict: `OK` - Findings: none ## Policy notes - GLM 5.1 + DeepSeek V4 Pro are the operator-required model mix for this bot. - Optional red-team model is enabled only when `PLATFORMCTL_PR_SANITY_REDTEAM_MODEL` is configured. - Auto-merge is not enabled here.