lulu-approach
Skill3 starsChanged 7 days agoMITView on GitHub
Domain holder for technical diagnostic decisions. It orchestrates one decision session under approach constraints; scripts own transitions, validation, and persistence. - $SKILL_ROOT, $WORKFLOW_DIR, $PLATFORM, $CACHE_DIR from ## Platform Context - $SKILL_DIR = $SKILL_ROOT/lulu-approach - $DECISION_SKILL_DIR = $SKILL_ROOT/decision - $APPROACH_ROOT = $CACHE_DIR/<cycle_id>/lulu-approach (the decision session) - Feature identification logic from ## Session Foundation </HARD-GATE> Decision macros ($DEC_START, $GATE_CONTROL, $RS_COMMIT, ...) are defined in ../decision/SKILL.md; <constraints_path> is $SKILL_DIR/constraints-$CYCLE_TYPE.json. Steps that Start, Resume, and Reopen invoke by name. Run before $DEC_START, $APPROACH_SESSION enter,…
In this repository11 other files it gives its agents.
- composeSkill
- decisionSkill
- eval-librarySkill
- lulu-betSkill
- lulu-brainstormSkill
- lulu-designSkill
- lulu-execSkill
- lulu-planSkill
- lulu-specSkill
- lulu-tasksSkill
- lulu-workflowSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
20 rules. About 0 are common, 7 are distinctive, and 13 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
13
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Do NOT proceed until you have read ../_runtime.md and ../decision/SKILL.md in full, and loaded:
- Very specific: line 43
Run before $DEC_START, $APPROACH_SESSION enter, or $APPROACH_SESSION reopen.
- Very specific: line 50
Run the docs macro. Parse stdout for files.
- Very specific: line 51
Read each path in files once.
- Very specific: line 56
Run after $DEC_START, $APPROACH_SESSION enter, or $APPROACH_SESSION reopen returns context_docs:
- Very specific: line 58
Run $GATE_CONTROL resolve-context and pin $CTX.
- Very specific: line 59
Read each returned context_docs path once.
- Distinctive: line 62
Do not re-run Pre-entry loading.
- Very specific: line 73
Run $RESOLVE_CONTEXT with --session-dir "$APPROACH_ROOT".
- Distinctive: line 83
If it reports a blocked prior stage, report that stage and do not retry.
- Very specific: line 85
Run the delegated DDF on Active.
- Distinctive: line 87
When the session is Completed, continue with Deliver.
- Distinctive: line 91
When Active remains this session:
- Very specific: line 104
Run $APPROACH_SESSION reopen.
- Very specific: line 106
Run $DEC_REOPEN --permit "$PERMIT_PATH", perform RS dialogue, then run $RS_COMMIT --gate "<G>" --operations '<json array>'.
- Very specific: line 116
Run $APPROACH_DELIVER. Session Completed authorizes this deliver.
- Distinctive: line 122
When any route reports an unresolved binding:
- Distinctive: line 124
Stop and report the binding ID and state from stderr.
- Very specific: line 125
Wait for an explicit human recovery decision.
- Distinctive: line 126
After external recovery, restart the applicable route and obtain fresh command output.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
lulu-approach
Domain holder for technical diagnostic decisions. It orchestrates one decision session under approach constraints; scripts own transitions, validation, and persistence.
Prerequisites
<HARD-GATE>
Do NOT proceed until you have read `../_runtime.md` and `../decision/SKILL.md` in full, and loaded:$SKILL_ROOT,$WORKFLOW_DIR,$PLATFORM,$CACHE_DIRfrom## Platform Context$SKILL_DIR=$SKILL_ROOT/lulu-approach$DECISION_SKILL_DIR=$SKILL_ROOT/decision$APPROACH_ROOT=$CACHE_DIR/<cycle_id>/lulu-approach(the decision session)- Feature identification logic from
## Session Foundation</HARD-GATE>
Script Macros
| Macro | Command |
$RESOLVE_CONTEXT | python3 "$SKILL_DIR/scripts/resolve_context.py" --cycle-id "<cycle_id>" --constraints "$SKILL_DIR/constraints-$CYCLE_TYPE.json" [--session-dir "<approach_root>"] |
$RESOLVE_CONTEXT_DOCS | python3 "$SKILL_DIR/scripts/resolve_context_docs.py" --cycle-id "<cycle_id>" --constraints "$SKILL_DIR/constraints-$CYCLE_TYPE.json" |
$RESOLVE_CONSTRAINT_DOCS | python3 "$SKILL_DIR/scripts/resolve_constraint_docs.py" --cycle-id "<cycle_id>" --constraints "$SKILL_DIR/constraints-$CYCLE_TYPE.json" |
$APPROACH_SHELL | python3 "$SKILL_DIR/scripts/approach_shell_control.py" --approach-root "$APPROACH_ROOT" |
$APPROACH_SESSION | python3 "$SKILL_DIR/scripts/approach_shell_control.py" --approach-root "$APPROACH_ROOT" <subcommand> --cycle-id "<cycle_id>" --constraints "$SKILL_DIR/constraints-$CYCLE_TYPE.json" |
$APPROACH_DELIVER | python3 "$SKILL_DIR/scripts/approach_shell_control.py" --approach-root "$APPROACH_ROOT" deliver --cycle-id "<cycle_id>" --confirm |
Decision macros ($DEC_START, $GATE_CONTROL, $RS_COMMIT, ...) are defined in ../decision/SKILL.md; <constraints_path> is $SKILL_DIR/constraints-$CYCLE_TYPE.json.
- Subcommand and stdout contracts remain in script module docstrings or
--help. - Any command that exits non-zero stops the flow: report stderr.
Shared procedures
Steps that Start, Resume, and Reopen invoke by name.
Pre-entry loading
Run before $DEC_START, $APPROACH_SESSION enter, or $APPROACH_SESSION reopen. For each kind, in this order:
| Kind | Docs macro | Rules |
| Context | $RESOLVE_CONTEXT_DOCS | $SKILL_DIR/references/context-rules.md |
| Constraint | $RESOLVE_CONSTRAINT_DOCS | $SKILL_DIR/references/constraint-rules.md |
- Run the docs macro. Parse stdout for
files. - Read each path in
filesonce. Iffilesis empty, skip reading. - Load the rules file and apply its sections in order (Principle → materials → Obligation → Body entry → Self-check). Do not classify documents into kinds yourself.
Context activation
Run after $DEC_START, $APPROACH_SESSION enter, or $APPROACH_SESSION reopen returns context_docs:
- Run
$GATE_CONTROL resolve-contextand pin$CTX. - Read each returned
context_docspath once. - Declare the active session to the user. Use only the newly pinned
$CTX.
Do not re-run Pre-entry loading; it runs only before session entry.
Routes
Resume and Reopen require the approach shell in Session state. From PackageReady or Delivered, report an unsupported state and stop.
Start
Once Session Foundation is complete:
- Complete Pre-entry loading.
- Run
$RESOLVE_CONTEXTwith--session-dir "$APPROACH_ROOT"; capture stdout path as$RESOLVED_CONTEXT_PATH. - Run
$APPROACH_SHELL init-shell. Run
$DEC_STARTwith:--constraints "$SKILL_DIR/constraints-$CYCLE_TYPE.json" \ --domain-constraints-file "$RESOLVED_CONTEXT_PATH" \ --session-dir "$APPROACH_ROOT"If it reports a blocked prior stage, report that stage and do not retry.
- Complete Context activation.
- Run the delegated DDF on Active. DC close includes
$GATE_CONTROL complete.
When the session is Completed, continue with Deliver.
Resume
When Active remains this session:
- Complete Pre-entry loading.
- Run
$APPROACH_SESSION enter. - Complete Context activation.
Then resume the delegated DDF on Active.
Reopen
Requires the approach shell in Session state. If the state is unknown, stop and report rather than infer it from files.
- Complete Pre-entry loading.
- Run
$APPROACH_SESSION reopen. Capture stdoutpermit_pathas$PERMIT_PATHandtransaction_idas$TRANSACTION_ID. - Complete Context activation.
- Run
$DEC_REOPEN --permit "$PERMIT_PATH", perform RS dialogue, then run$RS_COMMIT --gate "<G>" --operations '<json array>'. - Run
$APPROACH_SESSION complete-reopen --transaction-id "$TRANSACTION_ID".
Once the session is repaired, resume the delegated DDF on Active. Do not enter Deliver before then.
Deliver
Once the session is Completed ($GATE_CONTROL complete), enter PackageReady before delivering; Completed is not stage Delivered ($APPROACH_DELIVER):
- Run
$APPROACH_SHELL enter-package-ready. - Run
$APPROACH_DELIVER. Session Completed authorizes this deliver.
On success, announce stage complete from stdout next_steps (join when non-empty).
Interrupted binding
When any route reports an unresolved binding:
- Stop and report the binding ID and state from stderr.
- Wait for an explicit human recovery decision; do not auto-select
compensate-activeorcancel. - After external recovery, restart the applicable route and obtain fresh command output; do not infer the restored target.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

