skilly. Buy ad slot
All skills
Community / AGENT SKILL

speckit-token-observability

KamiJeong/tech-bridge-book
0 installs 0 GitHub stars
0

Required token analysis workflow for Codex Spec Kit orchestration phases.

BEFORE YOU INSTALL

Understand the trade-offs.

SECURITY REVIEW

Not yet assessed

Review the original instructions and requested permissions before installing.

No security review is available for this catalog entry yet.

SKILL QUALITY

Not yet assessed

How clearly the skill guides your agent, how complete its workflow is, and how you can check the outcome.

No quality assessment is available for this catalog entry yet.

The full skill.

Original instructions from the publisher’s SKILL.md

# Spec Kit Token Observability

Token analysis is a required observability step for every completed Spec Kit phase
and every custom orchestration phase. It is read-only with respect to product
artifacts and must not modify `spec.md`, `plan.md`, `tasks.md`, implementation
code, or contracts except for explicitly requested documentation updates outside
token capture.

## Run Labels

Use this stable format:

```text
<date>-<epic-slug>-<slice-id>-<mode>
```

Example:

```text
2026-05-18-admin-backoffice-s01-gated
```

## Artifact Locations

Each run writes under:

```text
.specify/token-analysis/runs/<run-label>/
├── metadata.json
├── steps/<phase>.json
├── summary.json
├── report.md
└── quality-notes.md
```

Shared comparison and baseline directories:

```text
.specify/token-analysis/comparisons/
.specify/token-analysis/baselines/
```

## Required Phases

Custom orchestration phases:

- intake
- slicing
- scheduling
- integration
- review
- commit
- PR creation

Spec Kit phases:

- specify
- clarify
- checklist
- plan
- tasks
- analyze
- implement

## Capture Methods

Use `exact` when runtime usage is available directly. Use `parsed` when usage is
extracted from a CLI JSON response. Use `estimated` when exact data is unavailable
but input/output artifacts allow a reasonable approximation. Use `unavailable`
only when neither exact nor estimated data can be produced; include a clear note.

For standard token-analyzer-supported phases, use the token-analyzer command or
hook when available. Supported standard phases include `specify`, `plan`, `tasks`,
and `implement`.

For unsupported or optional phases, use
`.agents/skills/speckit-guards/scripts/record-token-step.sh` to write an estimated
or unavailable step record. Do not silently skip token analysis.

## Missing Extension Behavior

Before a full workflow starts, run
`.agents/skills/speckit-guards/scripts/guard-token-analyzer-available.sh`.

If token-analyzer is missing:

- `plan-only` and `gated`: continue only after a clear warning.
- `auto-implement`, `auto-pr`, or any run with `auto-commit` enabled: stop
  before automation unless the user explicitly approves continuing.

Recommended installation command:

```bash
specify extension add token-analyzer
```

## Token Budgets

If `.specify/token-analysis/token-budget.json` exists, report warning and blocking
threshold status. Warning thresholds never block. Blocking thresholds block only
when explicitly configured as non-null values. If no budget exists, proceed
without budget enforcement and warn on missing token data by default.

## Final Output And PR Body

Final workflow output and draft PR bodies must include:

- Run label
- Token report path
- Total tokens, exact or estimated
- Most expensive phase
- Missing token data
- Token budget status
- Notes, including any approval to continue without token-analyzer