impact-measurement

Impact Measurement

Measure the quantitative cost of adding or changing an artifact: token injection cost, payload size, file growth, and context window consumption. All measurements come from tool calls against actual files — never from memory or estimates without a shown source.

This skill is one research angle in a multi-angle technical research system. It runs independently alongside api-state and ecosystem-research angles and returns its findings to the calling technical-researcher orchestrator.

Workflow

Step 1 — Scope

Measurement target received from orchestrator:

$ARGUMENTS

Identify from the above: the specific files or artifacts to measure, what is being added or changed, and which cost dimensions apply (token injection, discovery payload size, file size growth, item count change).

Step 2 — Baseline measurement

Measure the current state of each relevant file or component using tool calls. Record for each:

• File size in characters — use Bash(wc -c <path>) for byte/char counts; Read is for content inspection, not counting (its line-number prefixes distort counts)
• Line count where relevant
• Item counts (e.g., number of registered tools, prompts, skills)
• Discovery response size estimates (e.g., MCP tool list payload)

Every measurement must come from a Read, Bash, or equivalent tool call. State the source for each value.

Step 3 — Delta measurement

For each artifact being added or changed, measure its size directly:

• If the artifact exists: measure it with a tool call
• If the artifact does not exist yet (pre-implementation): measure its specified content source (e.g., the SKILL.md or prompt text that will be injected), and label the measurement as [pre-implementation estimate — source: <what was measured>]

Step 4 — Token cost estimation

Convert character counts to token estimates. Show the formula explicitly on every row. Label all values as estimates.

Conservative: tokens ≈ chars ÷ 3.5
Average:      tokens ≈ chars ÷ 4