datadog/create-runtime-setting
Create a new RuntimeSetting that can be changed at runtime via `agent config set/get` and the config API
npx skills add https://github.com/DataDog/datadog-agent --skill create-runtime-setting
Create a new RuntimeSetting implementation for the Datadog Agent. RuntimeSettings are settings that can be read and changed at runtime via:
agent config get <setting>, agent config set <setting> <value>, agent config list-runtimeGET /config/{setting}, POST /config/{setting}Use AskUserQuestion to collect the following. If $ARGUMENTS provides the setting name, skip that question.
log_payloads, dogstatsd_stats): the name used to register and access the setting via the API./config/list-runtime).pkg/config/settings/) — Used by multiple agent services (agent, trace-agent, process-agent, etc.)cmd/agent/subcommands/run/internal/settings/) — Only used by the core agentdatadog.yaml config key this setting maps to (e.g. log_payloads, internal_profiling.enabled). Often the same as the setting name, but can differ.cmd/agent/subcommands/run/command.go)cmd/cluster-agent/subcommands/start/command.go)cmd/trace-agent/subcommands/run/command.go)cmd/process-agent/subcommands/run/command.go)cmd/security-agent/subcommands/runtime/command.go)cmd/system-probe/subcommands/run/command.go)cmd/dogstatsd/subcommands/start/command.go)Before writing any code, read the appropriate reference files to follow existing patterns exactly.
comp/core/settings/component.go to understand the RuntimeSetting methods.Glob with pattern pkg/config/settings/runtime_setting_*.go to list available examples, then read one that matches the desired type (boolean, integer, string, etc.).command.go files listed in Step 1.7 to see how settings are added to the Settings map.File naming convention: runtime_setting_<feature_name>.go
File location:
pkg/config/settings/runtime_setting_<feature>.gocmd/agent/subcommands/run/internal/settings/runtime_setting_<feature>.goCreate the implementation following the patterns from the reference file read in Step 2. Every RuntimeSetting needs:
ConfigKey string fieldNew<Name>RuntimeSetting() that sets the config keyDescription() — returns the human-readable descriptionHidden() — returns whether hidden from list-runtimeName() — returns the config keyGet(config) — reads the current value using the appropriate typed getterSet(config, v, source) — validates/converts the input value, then calls config.Set()Type conversion in Set(): for Boolean and Integer types, use the GetBool(v) / GetInt(v) helper functions from pkg/config/settings — these handle string-to-type conversion. For agent-specific settings, import the helpers via settings "github.com/DataDog/datadog-agent/pkg/config/settings".
Create a test file alongside the implementation: runtime_setting_<feature>_test.go
Follow the test patterns from the reference test file read in Step 2. The test should verify:
Name(), Description(), Hidden() return expected valuesGet returns the correct value from configSet with a valid value updates the configSet with a string representation works (e.g. "true"/"false" for bools)Set with an invalid value returns an errorFind the settings.Params provider in the appropriate command.go file(s) for each selected service (from Step 1.7). Add the new setting to the Settings map following the existing pattern in that file. The import alias convention and registration format are visible in the existing entries.
dda inv test --targets=<package_path>
dda inv linter.go
RuntimeSetting interface is defined in comp/core/settings/component.goGetBool and GetInt are in pkg/config/settings/runtime_setting.goSet methods receive a model.Source parameter for config source tracking — always pass it through to config.Set()/config/{setting_name} (GET to read, POST to write) and via the CLI: agent config get <setting>, agent config set <setting> <value>, agent config list-runtime/create-runtime-setting — Interactive: prompts for all details/create-runtime-setting my_new_setting — Pre-fills the setting name, prompts for the restTake datadog/create-runtime-setting from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.