Correction Memory
v1.1.0Makes agent corrections persistent and reusable. When you override, reject, or correct an agent's output, this skill logs the correction and automatically in...
⭐ 0· 427·0 current·0 all-time
bydp-del@donovanpankratz-del
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description match what is provided: an instruction-only skill plus a local JS module that logs corrections to workspace/memory/corrections/[AgentType].jsonl and builds a preamble for injection. No unrelated env vars, binaries, or remote installs are requested.
Instruction Scope
SKILL.md limits actions to copying the included template into $OPENCLAW_WORKSPACE/lib, optionally wiring it into spawn logic, and inspecting files under memory/corrections. It does not instruct reading other system files, exfiltrating data, or calling external endpoints.
Install Mechanism
No network downloads or package installs. The install is a local copy of the included template file into the workspace (cp), which is low risk and consistent with the described behavior. Running the module triggers a harmless smoke test that writes to a temp dir.
Credentials
The skill requests no environment variables, credentials, or config paths. It requires a workspaceRoot argument to operate (used to place/read files under memory/corrections), which is appropriate for the stated functionality.
Persistence & Privilege
The skill is not always-enabled and does not request elevated privileges. If present, agent-context-loader (from another skill) may auto-load correction-tracker.js and prepend corrections to spawned agent prompts — intended behavior but worth noting because it changes agent spawn context across sessions.
Assessment
This skill appears to do exactly what it says: store human corrections under $OPENCLAW_WORKSPACE/memory/corrections and prepend recent corrections to future agent prompts. Before installing: 1) Inspect the template file (already included) and confirm you’re comfortable with storing corrections as plain text under memory/corrections; corrections are free text and could include sensitive info if you or agents log it. 2) Confirm $OPENCLAW_WORKSPACE points to the intended workspace and check file permissions so other processes/users can't read the corrections if that’s a concern. 3) Be aware that if you also use an intent-engineering/agent-context-loader module, this file will be auto-loaded and will change spawn behavior across sessions. 4) The module’s smoke test creates a temp directory when run directly — that’s benign but expected. If you want additional assurance, run the smoke test in a safe environment and review memory/corrections files after creating a few test corrections.Like a lobster shell, security has layers — review code before you run it.
agent-forgetsvk97djt45mc4dxk1vvrb6y5s5s181t4kcagent-improvementvk97djt45mc4dxk1vvrb6y5s5s181t4kcagent-learningvk97djt45mc4dxk1vvrb6y5s5s181t4kcagent-memoryvk97djt45mc4dxk1vvrb6y5s5s181t4kcagent-teachingvk97djt45mc4dxk1vvrb6y5s5s181t4kccontextvk97djt45mc4dxk1vvrb6y5s5s181t4kccorrection-injectionvk97djt45mc4dxk1vvrb6y5s5s181t4kccorrection-replayvk97djt45mc4dxk1vvrb6y5s5s181t4kccorrectionsvk97djt45mc4dxk1vvrb6y5s5s181t4kccross-sessionvk97djt45mc4dxk1vvrb6y5s5s181t4kcfeedbackvk97djt45mc4dxk1vvrb6y5s5s181t4kcimprovementvk97djt45mc4dxk1vvrb6y5s5s181t4kclatestvk9724dh75syrhgw03hcmc1v4t981wgvglearningvk97djt45mc4dxk1vvrb6y5s5s181t4kcmemoryvk97djt45mc4dxk1vvrb6y5s5s181t4kcmistakesvk97djt45mc4dxk1vvrb6y5s5s181t4kcopenclawvk97djt45mc4dxk1vvrb6y5s5s181t4kcpersistentvk97djt45mc4dxk1vvrb6y5s5s181t4kcremembervk97djt45mc4dxk1vvrb6y5s5s181t4kcrepeated-errorsvk97djt45mc4dxk1vvrb6y5s5s181t4kcspawnvk97djt45mc4dxk1vvrb6y5s5s181t4kcsubagentvk97djt45mc4dxk1vvrb6y5s5s181t4kcteachvk97djt45mc4dxk1vvrb6y5s5s181t4kc
License
MIT-0
Free to use, modify, and redistribute. No attribution required.