SuperAPI AI Directory

ecc-skill-defer

General & other

View repository

Updated 10-15 19:58

About this skill

/ecc-skill-defer — ECC Skill Progressive Loading Management

Manage the defer/restore state of ECC skills to reduce init token consumption.

Instructions

Run the script at this skill's directory: ecc-skill-defer/ecc-skill-defer.sh (relative to the skills repo root), or the installed copy at ~/.claude/scripts/ecc-skill-defer.sh.

Available commands

Command Description
status Show active/deferred counts
list Show all skills with their status
apply Defer skills listed in config (use after ECC update)
restore <name> Activate a single deferred skill
restore --all Restore all deferred skills
apply --reason "<text>" Defer with explicit reason logged to DEFER_LOG.md

Default behavior

If no argument is provided by the user, run status then list.

Config

The defer list is at ecc-skill-defer.conf (same directory as the script). To add/remove skills from the defer list, edit this file then run apply.

After restoring a skill

Tell the user: "Skill restored. It will be available in the next session (or after context reload)."

Notes

  • Defer reason tracking: apply --reason will record the defer reason in DEFER_LOG.md(append-only), with the format <date> | <skill> | <reason>。list The command can optionally display the most recent defer reason.
  • Portfolio health check: After a defer/restore operation, it is recommended to run /skill-health to check whether the current skill portfolio has redundancy or gaps.

Core skill protection (HITL Guard)

Before executing apply , you must first scan the list of skills about to be deferred (ecc-skill-defer.conf):

If the list contains design or assist, use AskUserQuestion to force confirmation (this is Hard Stop, and must not be skipped):

The list about to be deferred contains a core planning skill:

  • [list detected core skills]

design / assist are core planning skills; after deferring them, industry/academic reference capability and ECC resource allocation quality will be lost.

  1. Confirm defer — I understand the impact and still want to continue
  2. Cancel — keep these skills and do not execute apply

Only continue executing apply after the user explicitly selects "Confirm defer".

apply operation verification (basis: DAMA-DMBOK Completeness — manifest vs. actual)

Execute apply After that, you must re-verify the status via list :

  1. Before executing apply, record the list of skills expected to be deferred (manifest)

  2. After apply completes, run list to obtain the current status

  3. Verify item by item against the manifest (basis: OpenAI Developer Community consensus — Checklist-driven):

    Skill Expected status Actual status Result
    foo deferred deferred PASS
    bar deferred active FAIL
  4. If there is any FAIL → output a warning and prompt manual checking of ecc-skill-defer.conf and the script execution result

  5. All PASS → output the current list of available skills to ensure the user understands the existing resources

restore operation verification (basis: DAMA-DMBOK Completeness)

Execute restore <name> or restore --all After:

  1. Execute list to verify that the target skill status is active
  2. Output the verification result:[PASS] <skill> 已恢復為 active or [FAIL] <skill> 狀態仍為 deferred
  3. Inform the user: "Skill restored. It will be available in the next session (or after context reload).」